That said, trademark laws like life of the author + 95 years are absolutely absurd. The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property. The reasoning being that if you don't allow people to exclude 3rd party copying, then the primary party will assumedly not receive compensation for their creation and they'll never create.
Even in the case where the above is assumed true, the length of time that a protection should be afforded should be no more than the length of time necessary to ensure that creators create.
There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years. I wouldn't be surprised if it was the same for 1 year past death.
For that matter, this argument extends to other criminal penalties, but that's a whole other subject.
> The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property.
That was the original purpose. It has since been coopted by people and corporations whose incentives are to make as much money as possible by monopolizing valuable intangible "property" for as long as they can.
And the chief strategic move these people have made is to convince the average person that ideas are in fact property. That the first person to think something and write it down rightfully "owns" that thought, and that others who express it or share it are not merely infringing copyright, they are "stealing."
This plan has largely worked, and now the average person speaks and thinks in these terms, and feels it in their bones.
Trademark isn't copyright, those are two different things. Trademarks can be renewed roughly every 10 years [1] until the end of time and are about protecting a brand. Now copyright law lasts for "author plus 70 years. For anonymous works, pseudonymous works, or works made for hire, the copyright term is 95 years from the year of first publication or 120 years from creation, whichever comes first." [2]
Is copyright too long? Yes. Is it only that long to protect large media companies? Yes. But I would argue that AI companies are pushing the limits of fair use if not violating fair use, which is used as a affirmative defense by the way meaning that AI companies have to go to court to argue what they are doing is okay. They don't just get to wave their hands and say everything is okay because what we're doing is fair use and we get to scrape the world's entire creative output for our own profit.
> There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years.
I’m sure you’re right for individual authors who are driven by a creative spark, but for, say, movies made by large studios, the length of copyright is directly tied to the value of the movie as an asset.
If that asset generates revenue for 120 years, then it’s slightly more valuable than an asset that generates revenue for 119 years, and considerably more valuable than an asset that generates revenue for 20 years.
The value of the asset is in turn directly linked to how much the studio is willing to pay for that asset. They will invest more money in a film they can milk for 120 years than one that goes public domain after 20.
Would studios be willing to invest $200m+ in movie projects if their revenue was curtailed by a shorter copyright term? I don’t know. Probably yes, if we were talking about 120->70. But 120->20? Maybe not.
A dramatic shortening of copyright terms is something of a referendum on whether we want big-budget IP to exist.
In a world of 20 year copyright, we would probably still have the LOTR books, but we probably wouldn’t have the LOTR movies.
While I think the laws are broken, I also get why companies fight so hard to defend their IP: it is valuable, and they've built empires around it. But at some point, we have to ask: are we preserving culture or just hoarding it?
Regardless, it's not just copyright laws that are at issue here. This is reproducing human likenesses - like Harrison Ford's - and integrating them into new works.
So if I want to make an ad for a soap company, and I get an AI to reproduce a likeness of Harrison Ford, does that mean I can use that likeness in my soap commercials without paying him? I can imagine any court asking "how is this not simply laundering someone's likeness through a third party which claims to not have an image / filter / app / artist reproducing my client's likeness?"
All seemingly complicated scams come down to a very basic, obvious, even primitive grift. Someone somewhere in a regulatory capacity is either fooled or paid into accepting that no crime was committed. It's just that simple. This, however, is so glaring that even a child could understand the illegality of it. I'm looking forward to all of Hollywood joining the cause against the rampant abuse of IP by Silicon Valley. I think there are legal grounds here to force all of these models to be taken offline.
Additionally, "guardrails" that prevent 1:1 copies of film stills from being reprinted are clearly not only insufficient, they are evidence that the pirates in this case seek to obscure the nature of their piracy. They are the evidence that generative AI is not much more than a copyright laundering scheme, and the obsession with these guardrails is evidence of conspiracy, not some kind of public good.
Trademarks are very different from copyrights. In general, they never expire as long as the fees are paid in the region, and products bearing the mark are still manufactured. Note, legal firms will usually advise people that one can't camp on Trademarks like other areas of IP law.
For example, Aspirin is known as an adult dose of acetylsalicylic acid by almost every consumer, and is Trademarked to prevent some clandestine chemist in their garage making a similarly branded harmful/ineffective substance that damages the Goodwill Bayer earned with customers over decades of business.
Despite popular entertainment folklore, sustainable businesses actually want consumer goodwill associated with their products and services.
While I agree in many WIPO countries copyright law has essentially degenerated into monetized censorship by removing the proof of fiscal damages criteria (like selling works you don't own.) However, Trademarks ensure your corporate mark or product name is not hijacked in local Markets for questionable purposes.
Every castle needs a moat, but I do kind of want to know more about the presidential "Tesler". lol =)
You are missing a bunch of edge cases, and the law is all about edge cases.
An artist who works professionally has family members, family members who are dependent on them.
If they pass young, become popular just before they pass and their extremely popular works are now public domain. Their family sees nothing from their work, that is absolutely being commercialized ( publishing and creation generally spawns two seperate copyrights).
For what it's worth, this is a uniquely American view of copyright:
> The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property.
In Europe, particularly France, copyright arose for a very different reason: to protect an author's moral rights as the creator of the work. It was seen as immoral to allow someone's work -- their intellectual offspring -- to be meddled with by others without their permission. Your work represents you and your reputation, and for others to redistribute it is an insult to your dignity.
That is why copyrights in Europe started with much longer durations than they did in the United States, and the US has gradually caught up. It is not entirely a Disney effect, but a fundamental difference in the purpose of copyright.
> There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years. I wouldn't be surprised if it was the same for 1 year past death.
I think life of creator + some reasonable approximation of family members life expectancy would make sense. Content creators do create to ensure their family's security in some cases, I would guess.
You’re thinking of copyright law, not trademark law. Which serves a different function. If you’re going to critique something it’s useful to get your facts right.
Really? Because there are a lot of very stupid laws out there that should absolutely be broken, regularly, flagrantly, by as many people as possible for the sake of making enforcement completely null and pointless. Why write in neutered corporate-speak while commenting on a casual comment thread while also (correctly) pointing out the absurdity of certain laws.
this reminds me of the time I tried to use a prepaid lawyer to research some copyright issues
they went down the rabbit hole on trademark laws, which are not only not copyright related, they are an entirely different federal agency at the Patent Office
gave me a giggle and the last time I used cheapo prepaid lawyers
I honestly don't see why anyone who isn't JK Rowling should be allowed to coopt her world. I probably feel even more strongly for worlds/characters I like.
New rule: you get to keep your belongings for 20 years and that's it. Then they are taken away. Every dollar you made twenty or more years ago, every asset you acquired, gone.
That oughtta be enough to incentivize people to work and build their wealth.
Not sure if anyone is interested in this story, but I remember at the height of the PokemonGo craze I noticed there were no shirts for the different factions in the game, cant rememebr what they were called but something like Teamread or something. I setup an online shop to just to sell a red shirt with the word on it. The next day my whole shop was taken offline for potential copyright infringement.
What I found surprising is I didnt even have one sale. Somehow someone had notified Nintendo AND my shop had been taken down, to sell merch that didn't even exist for the market and if I remember correctly - also it didnt even have any imagery on it or anything trademarkable - even if it was clearly meant for pokmeonGo fans.
Im not bitter I just found it interesting how quick and ruthless they were. Like bros I didn't even get a chance to make a sale. ( yes and also I dont think I infringed anything).
I asked Sora to turn a random image of my friend and myself into Italian plumbers. Nothing more, just the two words "Italian plumbers". The created picture was not shown to me because it was in violation of OpenAI's content policy. I asked then just to turn the guys on the picture into plumbers, but I asked this in the Italian language.
Without me asking for it, Sora put me in an overall and gave me a baseball cap, and my friend another baseball cap. If I asked Sora to put mustache on us, one of us received a red shirt as well, without being asked to. Starting with the same pic, if I asked to put one letter on the baseball caps each - guess, the letters chosen were M and L.
These extra guardrails are not really useful with such a strong, built-in bias towards copyright infringement of these image creation tools.
Should it mean that with time, Dutch pictures will have to include tulips, Italian plumbers will have to have a uniform with baseball caps with L and M, etc. just not to confuse AI tools?
Many years ago I tried to order a t-shirt with the postscript tiger on the front from Spreadshirt.
It was removed on Copyright claims before I could order one item myself. After some back and forth they restored it for a day and let me buy one item for personal use.
My point is: Doesn't have to be Sony, doesn't have to be a snitch - overzealous anticipatory obedience by the shop might have been enough.
Twenty years ago, I worked for Google AdWords as a customer service rep. This was still relatively early days, and all ads still had some level of manual human review.
The big advertisers had all furnished us a list of their trademarks and acceptable domains. Any advertiser trying to use one that wasn’t on the allow-list had their ad removed at review time.
I suspect this could be what happened to you. If the platform you were using has any kind of review process for new shops, you may have run afoul of pre-registered keywords.
Well the teams in Pokemon Go aren't quite as generic as Teamred: they are Team Instinct, Team Mystic, and Team Valor. Presumably Nintendo has trademarks on those phrases, and I’m sure all the big print on demand houses have an API for rights-holders to preemptively submit their trademarks for takedowns.
Nintendo is also famously protective of their IP: to give another anecdote, I just bought one of the emulator handhelds on Aliexpress that are all the rage these days, and while they don't advertise it they usually come preloaded with a buttload or ROMs. Mine did, including a number of Nintendo properties — but nary an Italian plumber to be found. The Nintendo fear runs deep.
Allen Pan, a youtuber "maker" who runs in the circle of people who run OpenSauce, was a contestant on a Discovery channel show that was trying to force the success of Mythbusters by "finding the next mythbusters!". He lost, but it was formative to him because those people were basically all inspired by the original show.
A couple years ago, he noticed that the merchandise trademark for "Mythbusters" had lapsed, so he bought it. He, now the legal owner of the trademark Mythbusters for apparel, made shirts that used that trademark.
Discovery sent him a cease and desist and threatened to sue. THEY had let the trademark lapse. THEY had lost the right to the trademark, by law. THEY were in the wrong, and a lawyer agreed.
But good fucking luck funding that legal battle. So he relinquished the trademark.
Hilariously, a friend of Allen Pan's, from the same "Finding the next mythbuster" show; Kyle Hill, is friends enough with Adam Savage to talk to him occasionally, and supposedly the actual Mythbusters themselves were not empathetic to Allen's trademark claim.
I think the problem there was being dependent on someone who is a complete pushover, doesn't bother to check for false positives and can kill your business with a single thought.
Idk, the models generating what are basically 1:1 copies of the training data from pretty generic descriptions feels like a severe case of overfitting to me. What use is a generational model that just regurgitates the input?
I feel like the less advanced generations, maybe even because of their limitations in terms of size, were better at coming up with something that at least feels new.
In the end, other than for copyright-washing, why wouldn't I just use the original movie still/photo in the first place?
People like what they already know. When they prompt something and get a realistic looking Indiana Jones, they're probably happy about it.
To me, this article is further proof that LLMs are a form of lossy storage. People attribute special quality to the loss (the image isn't wrong, it's just got different "features" that got inserted) but at this point there's not a lot distinguishing a seed+prompt file+model from a lossy archive of media, be it text or images, and in the future likely video as well.
The craziest thing is that AI seems to have gathered some kind of special status that earlier forms of digital reproduction didn't have (even though those 64kbps MP3s from napster were far from perfect reproductions), probably because now it's done by large corporations rather than individuals.
If we're accepting AI-washing of copyright, we might as well accept pirated movies, as those are re-encoded from original high-resolution originals as well.
Tried Flux.dev with the same prompts [0] and it seems actually to be a GPT problem. Could be that in GPT the text encoder understands the prompt better and just generates the implied IP, or could be that a diffusion model is just inherently less prone to overfitting than a multimodal transformer model.
If it overfits on the whole internet then it’s like a search engine that returns really relevant results with some lossy side effect.
Recent benchmark on unseen 2025 Math Olympiad shows none of the models can problem solve . They all accidentally or on purpose had prior solutions in the training set.
What if the word "generic" were added to a lot of these image prompts? "generic image of an intergalactic bounty hunter from space" etc.
Certainly there's an aspect of people using the chat interface like they use google: describe xyz to try to surface the name of a movie. Just in this case, we're doing the (less common?) query of: find me the picture I can vaguely describe; but it's a query to a image /generating/ service, not an image search service.
Idk, a couple of the examples might be generic enough that you wouldn't expect a very specific movie character. But most of the prompts make it extremely clear which movie character you would expect to see, and I would argue that the chat bot is working as expected by providing that.
Yeah, I've been feeling the same. When a model spits out something that looks exactly like a frame from a movie just because I typed a generic prompt, it stops feeling like “generative” AI and more like "copy-paste but with vibes."
> I feel like the less advanced generations, maybe even because of their limitations in terms of size, were better at coming up with something that at least feels new.
Ironically that's probably because the errors and flaws in those generations at least made them different from what they were attempting to rip off.
I'm not sure if this is a problem with overfitting. I'm ok with the model knowing what Indiana Jones or the Predator looks like with well remembered details, it just seems that it's generating images from that knowledge in cases where that isn't appropriate.
I wonder if it's a fine tuning issue where people have overly provided archetypes of the thing that they were training towards. That would be the fastest way for the model to learn the idea but it may also mean the model has implicitly learned to provide not just an instance of a thing but a known archetype of a thing. I'm guessing in most RLHF tests archetypes (regardless of IP status) score quite highly.
Probably an over-representation in the training data really so it's causing overfitting. Because using training data in amounts right from the Internet it's going to be opinionated on human culture (Bart Simpson is popular so there are lots of images of him, Ori is less well known so there are fewer images). Ideally it should be training 1:1 for everything but that would involve _so_ much work pruning the training data to have a roughly equal effect between categories.
One thing I would say, it's interesting to consider what would make this not so obviously bad.
Like, we could ask AI to assess the physical attributes of the characters it generated. Then ask it to permute some of those attributes. Generate some random tweaks: ok but brawy, short, and a different descent. Do similarly on some clothing colors. Change the game. Hit the "random character" button on the physical attributes a couple times.
There was an equally shatteringly-awful less-IP-theft (and as someone who thinks IP is itself incredibly ripping off humanity & should be vastly scoped down, it's important to me to not rest my arguments on IP violations).... An equally shattering recent incident for me. Having trouble finding it, don't remember the right keywords, but an article about how AI has a "default guy" type that it uses everywhere, a super generic personage, that it would use repeatedly. It was so distasteful.
The nature of 'AI as compression', as giving you the most median answer is horrific. Maybe maybe maybe we can escape some of this trap by iterating to different permutations, by injecting deliberate exploration of the state spaces. But I still fear AI, worry horribly when anyone relies on it for decision making, as it is anti-intelligent, uncreative in extreme, requiring human ingenuity to budge off its rock of oppressive hypernormality that it regurgitates.
Are you telling me that our culture should be deprived of the idea of Indiana Jones and the feelings that character inspires in all of us forever just because a corporation owns the asset?
Indiana Jones is 44 years old. When are we allowed to remix, recreate and expand on this like humanity has done since humans first started sitting down next to a fire and telling stories?
Mandrake: Colonel... that Coca-Cola machine. I want you to shoot the lock off it. There may be some change in there.
Guano: That's private property.
Mandrake: Colonel! Can you possibly imagine what is going to happen to you, your frame, outlook, way of life, and everything, when they learn that you have obstructed a telephone call to the President of the United States? Can you imagine? Shoot it off! Shoot! With a gun! That's what the bullets are for, you twit!
Guano: Okay. I'm gonna get your money for ya. But if you don't get the President of the United States on that phone, you know what's gonna happen to you?
Mandrake: What?
Guano: You're gonna have to answer to the Coca-Cola company.
I guess we all have to answer to the Walt Disney company.
But I can hire an artist and ask him to draw me a picture of Indiana Jones, he creates a perfect copy and I hang it on my fridge. Where did I (or the artist) violate any copyright (or other) laws? It is the artist that is replaced by the AI, not the copyrighted IP.
I mean... If I go to Google right now and do an image search for "archeologist adventurer who wears a hat and uses a bullwhip," the first picture is a not-even-changed image of Indiana Jones. Which I will then copy and paste into whatever project I'm working on without clicking through to the source page (usually because the source page is an ad-ridden mess).
Perhaps the Internet itself is the hideous theft machine, and AI is just the most efficient permutation of user interface onto it.
(Incidentally, if you do that search, you will also, hilariously, turn up images of an older gentleman dressed in a brown coat and hat who is clearly meant to be "The Indiana Jones you got on Wish" from a photo-licensing site. The entire exercise of trying to extract wealth via exclusive access to memetic constructs is a fraught one).
I hate how it is common to advance a position to just state a conclusion as if it were a fact. You keep repeating the same thing over and over until it seems like a concensus has been reached instead of an actual argument reasoned from first principle.
This is no theft here. Any copyright would be flimsier than software patents. I love Studio Ghibli (including $500/seat festival tickets) but it's the heart and the detail that make them what they are. You cannot clone that. Just some surface similarity. If that's all you like about the movies... you really missed the point.
Imagine if in early cinema someone had tried to claim mustachioed villian, ditsy blonde, or dumb jock? These are just tropes and styles. Quality work goes much much much deeper, and that cannot be synthesised. I can AI generate a million engagement rings, but I cannot pick the perfect one that fits you and your partners love story.
PS- the best work they did was "When Marnie was There". Just fitted together perfectly.
Interesting proposal. Maybe if race or sex or height or eye color etc isn't given, and the LLM determines there's no reason not to randomize in this case (avoid black founding fathers), the backend could tweak its own prompt by programatically inserting a few random traits to the prompt.
If you describe an Indiana Jones character, but no sex, 50/50 via internal call to rand() that it outputs a woman.
Ah, I thought I knew this account from somewhere. It seems surprisingly easy to figure out what account is commenting just based on the words used, as I've commented that only a few active people on this site seem to use such strong words as shown here.
So if it’s a theft machine, how is the answer to try teaching it to hide the fact that it’s stealing by changing its outputs? That’s like a student plagiarizing an essay and then swapping some words with a thesaurus pretending that changes anything.
Wouldn’t the more appropriate solution in the case of theft be to remunerate the victims and prevent recidivism?
Instead of making it “not so obviously bad” why not just… make it good? Require AI services to either prove that 100% of their training corpus is either copyright free or properly licensed, or require them to compensate copyright holders for any infringing outputs.
Looks to me like OpenAI drew their guardrails somewhere along a financial line. Generate a Micky Mouse or a Pikachu? Disney and Pokemon will sue the sh*t out of you. Ghibli? Probably not powerful enough to risk a multimillion years long court battle.
I think the cat is out of the bag when it comes to generative AI, the same way how various LLMs for programming have been trained even on codebases that they had no business using, yet nobody hasn’t and won’t stop them. It’s the same as what’s going to happen with deepfakes and such, as the technology inevitably gets better.
> Hayao Miyazaki’s Japanese animation company, Studio Ghibli, produces beautiful and famously labor intensive movies, with one 4 second sequence purportedly taking over a year to make.
It makes me wonder though - whether it’s more valuable to spend a year on a scene that most people won’t pay that much attention to (artists will understand and appreciate, maybe pause and rewind and replay and examine the details, the casual viewer just enjoy at a glance) or use tools in addition to your own skills to knock it out of the park in a month and make more great things.
A bit how digital art has clear advantages over paper, while many revere the traditional art a lot, despite it taking longer and being harder. The same way how someone who uses those AI assisted programming tools can improve their productivity by getting rid of some of the boilerplate or automate some refactoring and such.
AI will definitely cheapen the art of doing things the old way, but that’s the reality of it, no matter how much the artists dislike it. Some will probably adapt and employ new workflows, others stick to tradition.
It's a very clear difference between a cheap animation and Ghibli. Anyone can see it.
In the first case, there's only one static image for an entire scene, scrolled and zoomed, and if they feel generous, there would be an overlay with another static image that slides over the first at a constant speed and direction. It feels dead.
In the second case, each frame is different. There's chaotic motions such as wind and there's character movement with a purpose, even in the background, there's always something happening in the animation, there's life.
Fundamentally I think this comes down to answering the question of "why are you creating this?".
There are many valid answers.
Maybe you want to create it to tell a story, and you have an overflowing list of stories you're desperate to tell. The animation may be a means to an end, and tools that help you get there sooner mean telling more stories.
Maybe you're pretty good at making things people like and you're in it for the money. That's fine, there are worse ways to provide for your family than making things people enjoy but aren't a deep thing for you.
Maybe you're in it because you love the act of creating it. Selling it is almost incidental, and the joy you get from it comes down to spending huge amounts of time obsessing over tiny details. If you had a source of income and nobody ever saw your creations, you'd still be there making them.
These are all valid in my mind, and suggest different reasons to use or not to use tools. Same as many walks of life.
I'd get the weeds gone in my front lawn quickly if I paid someone to do it, but I quite enjoy pottering around on a sunny day pulling them up and looking back at the end to see what I've achieved. I bake worse bread than I could buy, and could buy more and better bread I'm sure if I used the time to do contracting instead. But I enjoy it.
On the other hand, there are things I just want done and so use tools or get others to do it for me.
One positive view of AI tools is that it widens the group of people who are able to achieve a particular quality, so it opens up the door for people who want to tell the story or build the app or whatever.
A negative side is the economics where it may be beneficial to have a worse result just because it's so much cheaper.
> the same way how various LLMs for programming have been trained even on codebases that they had no business using, yet nobody hasn’t and won’t stop them
I remember the discourse on HN a few years ago when GitHub Copilot came out, and well, people were missing the fact that the GitHub terms and conditions (T&C) explicitly permits usage of code for analytic purposes (of which training is implicitly part), so it turns out that if one did not want such training, they should not have hosted on GitHub in the first place, because the website T&C was clear as day.
>It makes me wonder though - whether it’s more valuable to spend a year on a scene that most people won’t pay that much attention to (artists will understand and appreciate, maybe pause and rewind and replay and examine the details, the casual viewer just enjoy at a glance) or use tools in addition to your own skills to knock it out of the park in a month and make more great things.
If they didn't spend a year on it they wouldn't be copied now.
So many arguing that "copyright shouldn't be a thing" etc., ad nauseam, which is a fine philosophical debate. But it's also the law. And that means ChatGPT et. al. also have to follow the law.
I really, really hope the multimedia-megacorps get together and class-action ChatGPT and every other closed, for-profit LLM corporation into oblivion.
There should not be a two-tier legal system. If it's illegal for me, it's illegal for Sam Altman.
That’s a fine philosophical debate, but the law is designed by the rich to favor the rich and while there are a number of exceptions there is little you can do with the legal system without money and lots of it. So while having a truly just system would be neat it just isn’t in the cards for humanity (IMHO) so long as we allow entities to amass “fuck you” money and wield it to their liking.
There is more to it than copyright when you start going down the path of photorealism. As much as it is a picture of Indiana jones, it is also a picture of Harrison Ford. As fun as it is to make hilarious videos of presidents sucking ceo toes, there has to be a line.
There is a lack of consent here that runs even deeper than what copyright was traditionally made to protect. It goes further than parody. We can't flip our standards back and forth depending on who the image is made to reproduce
Oooh those guardrails make me angry. I get why they are there (dont poke the bear) but it doesn't make me overlook the self serving hypocrisy involved.
Though I am also generally opposed to the notion of intellectual property whatsoever on the basis that it doesn't seem to serve its intended purpose and what good could be salvaged from its various systems can already be well represented with other existing legal concepts, i.e deceptive behaviors being prosecuted as forms of fraud.
The problem is people at large companies creating these AI models, wanting the freedom to copy artists’ works when using it, but these large companies also want to keep copyright protection intact, for their regular business activities. They want to eat the cake and have it too. And they are arguing for essentially eliminating copyright for their specific purpose and convenience, when copyright has virtually never been loosened for the public’s convenience, even when the exceptions the public asks for are often minor and laudable. If these companies were to argue that copyright should be eliminated because of this new technology, I might not object. But now that they come and ask… no, they pretend to already have, a copyright exception for their specific use, I will happily turn around and use their own copyright maximalist arguments against them.
It's not just the guardrails, but the ham-fisted implementation.
Grok is supposed to be "uncensored", but there are very specific words you just can't use when asking it to generate images. It'll just flat out refuse or give an error message during generation.
But, again, if you go in a roundabout way and avoid the specific terms you can still get what you want. So why bother?
Is it about not wanting bad PR or avoiding litigation?
Everyone is talking about theft - I get it, but there's a more subtler point being made here.
Current generation of AI models can't think of anything truly new. Everything is simply a blend of prior work. I am not saying that this doesn't have economic value, but it means these AI models are closer to lossy compression algorithms than they are to AGI.
The following quote by Sam Altman from about 5 years ago is interesting.
"We have made a soft promise to investors that once we build this sort-of generally intelligent system, basically we will ask it to figure out a way to generate an investment return."
That's a statement I wouldn't even dream about making today.
Disregarding the (common!) assumption that AGI will consist of one monolithic LLM instead of dozens of specialized ones, I think your comment fails to invoke an accurate, consistent picture of creativity/"truly new" cognition.
To borrow Chomsky's framework: what makes humans unique and special is our ability to produce an infinite range of outputs that nonetheless conform to a set of linguistic rules. When viewed in this light, human creativity necessarily depends on the "linguistic rules" part of that; without a framework of meaning to work within, we would just be generating entropy, not meaningful expressions.
Obviously this applies most directly to external language, but I hope it's clear how it indirectly applies to internal cognition and--as we're discussing here--visual art.
TL;DR: LLMs are definitely creative, otherwise they wouldn't be able to produce semantically-meaningful, context-appropriate language in the first place. For a more empirical argument, just ask yourself how a machine that can generate a poem or illustration depicting [CHARACTER_X] in [PLACE_Y] doing [ACTIVITY_Z] in [STYLE_S] without being creative!
I don't see why this is an issue? The prompts imply obvious and well-known characters, and don't make it clear that they want an original answer. Most humans would probably give you similar answers if you didn't add an additional qualifier like "not Indiana Jones". The only difference is that a human can't exactly reproduce the likeness of a famous character without significant time and effort.
The real issue here is that there's a whole host of implied context in human languages. On the one hand, we expect the machine to not spit out copyrighted or trademarked material, but on the other hand, there's a whole lot of cultural context and implied context that gets baked into these things during training.
I think the point is that for a lot of them there are endless possible alternatives to the character design, but it still generates one with the exact same design. Why can't, for example, the image of Tomb Raider have a different colored tank top? Why is she wearing a tank top and not a shirt? Why does she have to have a gun? Why is she a busty, attractive brunette? These are all things that could be different but the dominance of Lara Croft's image and strong association with the words "tomb raider" in popular culture clearly influences the model's output.
Normally (well, if you're ethical) credit is given.
Also, there are IP limits of various sorts (e.g. copyright, trademark) for various purposes (some arguably good, some arguably bad), and some freedoms (e.g., fair use). There's no issue if this follows the rules... but I don't see where that's implemented here.
It looks like they may be selling IP they don't own the right to.
> Maybe Studio Ghibli making it through the seemingly deterministic GPT guardrails was an OpenAI slip up, a mistake,
The author is so generous... but Sam Altman literally has a Ghibli-fied Social profile and in response to all this said OpenAI chooses its demos very carefully. His primary concern is that Ghibli-fying prompts are over-consuming their GPU resources, degrading the service by preventing other ChatGPT tasks.
The official White House account has been posting ghiblified images too, Altman knows that as long as he's not critical of the current administration he's untouchable.
I was really hoping that the conversation around AI art would at least be partially centered on the perhaps now dated "2008 pirate party" idea that intellectual property, the royalty system, the draconian copyright laws that we have today are deeply silly, rooted in a fiction, and used over and over again, primarily by the rich and powerful, to stifle original ideas and hold back cultural innovation.
Unfortunately, it's just the opposite. It seems most people have fully assimilated the idea that information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus. That Disney Corp can prevent you from viewing some collection of pixels, because THEY own it, and they know better than you do about the culture and communication that you are and are not allowed to experience.
It's just baffling. If they could, Disney would scan your brain to charge you a nickel every time you thought of Mickey Mouse.
The idea of open sourcing everything and nullifying patents would benefit corporations like Disney and OpenAI vastly more than it would benefit the people. The first thing that would happen is that BigCorp would eat up every interesting or useful piece of art, technology, and culture that has ever been created and monetize the life out of it.
These legal protections are needed by the people. To the Pirate Party's credit, undoing corporate personhood would be a good first step, so that we can focus on enforcing protections for the works of humans. Still, attributing those works to CEOs instead of corporations wouldn't result in much change.
If we are going to have a general discussion about copyright reform at a national level, I'm all for it. If we are going to let billion dollar corporations break the law to make even more money and invent legal fictions after the fact to protect them, I'm completely against it.
Training a model is not equivalent to training a human. Freedom of information for a mountain of graphics cards in a privately owned data center is not the same as freedom of information for flesh and blood human beings.
A different way of looking at it: AI, by design, defaults to regurgitating the poppiest of pop culture content. Every whip-wielding archaeologist is now Harrison Ford. Every suave British spy is now Daniel Craig. With the power of AI, creativity is dead and buried.
> That Disney Corp can prevent you from viewing some collection of pixels, because THEY own it
A world without copyright is just as problematic as a world with copyright. With copyright, you run into the problem of excessive control. This wasn't too much of a problem in the past. If you bought a book, record, or video recording, you owned that particular copy. You could run into disagreeable situations because you didn't own the rights, but it was difficult to prevent anyone from from viewing a work once it had been published. (Of course, modern copyright laws and digitial distribution has changed that.)
On the flip side, without copyright, it would be far easier for others to exploit (or even take credit) for the work of another person without compensation or recourse. Just look at those AI "generated" images, or any website that blatently rips off the content created by another person. There is no compensation. Heck, there isn't even credit. Worse yet, the parties misrepresenting the content are doing their best to monetize it. Even authors who are more than willing to give their work away have every right to feel exploited under those circumstances. And all of that is happening with copyright laws, where there is the opportunity for recourse if you have the means and the will.
I think what you observe is more like a natural blowback to the prevailing idea that this is somehow beyond critique because it will fundamentally change culture and civilization forever.
There's a bit of irony here too. The intellectual discourse around intellectural property, a diverse and lively one from an academic standpoint, the whole free and open source software movements, software patents, the piracy movement and so on have analyzed the history, underlying ideas and values in detail for the past thirty years. Most people know roughly what is at stake, where they stand, and can defend their position in an honest way.
Then comes new technology, everyone and their mother gets excited about it, and steamrolls all those lofty ideas into "oh look at all the shiny things it can produce!". Be careful what you wish for.
Not just some particular collection of pixels, but an infinite number of combinations of collections of pixels, any of which remotely invoke a shadow of similarity to hundreds of "properties" that Disney lays claim to.
But this must include the dissolution of patents. Otherwise corporations and the owners of the infrastructure will simply control everything, including the easily replicable works of individuals.
No matter the extent you believe in the freedom of information, few believe anyone should then be free to profit from someone else's work without attribution.
You seem to think it would be okay for disney to market and charge for my own personal original characters and art, claiming them as their own original idea. Why is that?
I'm guessing you've never created something of value before. People are entitled to the fruits of their labour and control of their intellectual property.
I think we have all grown up with pervasive strong IP rights, and most people have come to internalize it as a matter of fairness or an almost natural right, rather than a practical tool designed to incentivize creation.
And then even if you get past that, the world is filled with lots of IP we love, and it is easy to imagine weakened IP rights taking that away, but quite difficult to imagine what weakened IP rights might buy us.
I do have some hope still that this generative AI stuff will give a glimpse into the value of weaker IP rights and maybe inspire more people to think critically about it. But I think it is an uphill battle. Or maybe it will take younger people growing up on generative AI to notice.
Consider that one day you may wish to author a creative work and derive financial benefit from that labor. There is legitimate use for limited time ownership of reproducible cultural artifacts. Extending that to 95 years is the problem.
I can't speak for everyone obviously, but my anti-AI sentiment in this regard is not that IP law is flawless and beyond reproach, far from it. I'm merely saying that as long as we're all required to put up with it, that OpenAI and company should also have to put up with it. It's incredibly disingenuous the way these companies have taken advantage of publicly available material on an industrial scale, used said material to train their models "for research" and as soon as they had something that vaguely did what they wanted, began selling access to them.
If they are indeed the output of "research" that couldn't exist without the requisite publicly available material, then they should be accessible by the public (and arguably, the products of said outputs should also be inherently public domain too).
If they are instead created products to be sold themselves, then what is utilized to create them should be licensed for that purpose.
Additionally, if they can be used to generate IP violating material, then IMHO, makes perfect sense for the rights holders of those IPs to sue their asses like they would anyone else who did that and sold the results.
Again, for emphasis: I'm not endorsing any of the effects of IP law. I am simply saying that we should all, from the poorest user to the richest corporation, be playing by the same rules, and it feels like AI companies entire existence is hinging on their ability to have their IP cake and eat it too: they want to be able to restrict and monetize access to their generative models that they've created, while also having free reign to generate clearly, bluntly plagiarizing material, by way of utilizing vast amounts of in-good-faith freely given material. It's gross, and it sucks.
* If I make a thing that is different and I get a patent - cool.
* If I create a design that is unusual and I get copyright on it - is that cool?
Both concepts - patent and copyright - are somewhat controversial, for multiple reasons.
If you invented a thingie, would you not want some initial patent related protection to allow you to crack on with some sort of clout against cough CN? If you created a film/creative thang, would you not want some protection against your characters being ... subverted.
Patents and copywrite are what we have - do you have any better ideas?
What's the damage to the society done by Disney holding the rights to Mickey Mouse? Like, if we're being honest?
Patents, sure. They're abused and come at a cost to the society. But all we've done here is created a culture where, in some sort of an imagined David-vs-Goliath struggle against Disney, we've enabled a tech culture where it's OK to train gen AI tech on works of small-scale artists pilfered on an unprecedented scale. That's not hurting Disney. It's hurting your favorite indie band, a writer you like, etc.
We're about to witness a fundamental shift in the human experience. Some time in the near future there will not be a single act of creation you can do that isn't trivial compared to the result of typing "make cool thing please now" into the computer. And your position is to add to the problem because with your policy anything I create should get chucked into the LLM grinder by any and everybody. How do I get my human body to commit to doing hard things with that prospect at hand? This is the end of happiness.
> It seems most people have fully assimilated the idea that information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus.
No, the idea is that rules needed to be changed in a way that can are valid for everyone, not just for mega corporations who are trying to exploit other's works and gatekeep the it behind "AI".
> used over and over again, primarily by the rich and powerful
This is where the argument falls apart. Not because the copyright isn't used by the rich and powerful, but because it misses the fact that copyright also grants very powerful rights to otherwise powerless individuals, thus allowing for many small businesses and individuals to earn a living based on the rights granted by our current copyright system.
>information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus
I think that's the reason why I've (and probably many others?) have cooled down on general open source coding.
Open source started when well-paid programmers used their stable positions and ample extra time to give back to the community. What happened then is that corporations then siphoned up all that labor and gave nothing back, just like the AI bros siphoned up all data and gave nothing back. The 'contract' of mutual exchange, of bettering each other, was always a fantasy. Instead the companies took away that ample extra time and those stable positions.
Here we are in 2025 and sometimes I can't afford rent but the company C-tier is buying itself their hundredth yacht. Why should I contribute to your system?
Getting the megacorporations to sit up and take notice of this is about the only way the average independent artist has any hope of stopping this crap from destroying half our jobs. What'm I gonna do, sue OpenAI? Sam Altman makes more money sitting on the toilet taking a dump than I do in an entire year.
I have no love for the Mouse but if I can get them and the image slop-mongers to fight then that's absoutely fine. It would be nice to have a functioning, vibrant public domain but it is also nice to not have some rich asshole insisting that all copyright laws must be ignored because if they properly licensed even a fraction of what they've consumed then it would be entirely too expensive to train their glorified autocomplete databases on the entire fucking internet for the purpose of generating even more garbage "content" designed to keep your attention when you're mindlessly scrolling their attention farms, regardless of how it makes you feel, and if I can choose one or the other then I am totally behind the Mouse.
The problem with this kind of plagiarism isn't that it violates someone's specific copyright.
But the discussion around plagiarism calls attention to the deeper issue: "generative" AI does not have emergent thinking or reasoning capabilities. It is just very good at obfuscating the sources of its information.
And that can cause much bigger problems than just IP infringement. You could make a strategic decision based on information that was deliberately published by an adversary.
It smells like a psyop, to be honest. Doesn't take much to get the ball rolling. Just more temporarily embarrassed millionaires sticking up for billionaires and corporations, buying their propaganda hook line and sinker, and propagating it themselves for free. Copyright is a joke, DMCA is a disgusting, selectively applied tool of the elite.
All those ideas were rationalizations because people didn’t want to pay for stuff, just like your post effectively blaming the victim of IP theft cause corporations undeniably do suck so we shouldn’t care if they suffer.
> used over and over again, primarily by the rich and powerful, to stifle original ideas and hold back cultural innovation
You're using those "2008 ideas now to defend the rich and powerful exploiting and stifling creativity; the problem hasn't changed, you've just swapped sides.
I don't understand how protecting Disney characters prevents development of art or science. Why do you need them at all? There is lot of liberally licensed art and I think today there are more artists than ever in history.
Also making a billion dollar business by using hard work of talented people for free and without permission is not cool. The movie they downloaded from Pirate Bay for free took probably man-years of work to make.
Also I wonder how can we be sure that the images produced by machine are original and are not a mix of images from unknown artists at DeviantArt. Maybe it is time to make a neural image origin search engine?
I agree with the sentiment elsewhere in this thread that this represents a "hideous theft machine", but I think even if we discard that, this is still bad.
It's very clear that generative has abandoned the idea of creative; image production that just replicates the training data only serves to further flatten our idea of what the world should look like.
> I don’t know…the actual inspirations for Indiana Jones, like Allan Quatermain from H. Rider Haggard's novels, "King Solomon's Mines", and the real life Roy Chapman Andrews, who led expeditions to Mongolia and China in the 1920s and wore a fedora.
The actual inspiration for Indy was protagonist Harry Steele from the movie The Secret of the Incas (1954). Filmed on location in Cusco and Machu Picchu, before they became popular tourist destinations, the movie also had scenes and elements that made it into Raiders of the Lost Ark.
Definitely a missed opportunity that the article didn't discuss that obviously-derivative borrowing has been happening a lot longer than ML image generation has been around. And that borrowing is OK! Indiana Jones' appearance is very obviously based directly on Charlton Heston's character in Secret of the Incas, but the Spielberg/Lucas films are objectively better in every way than that source material.
I remember when google news was fined be the EU for just linking and using some preview + trailer text to actual news websites. The news are owned by a few monopolies and they don't like giving up control.
I received so many Copyright and DMCA takedowns for early youtube videos posted in the early 2010's for no reason except some background music blaring a hit. It had millions of views and NO ADs. Now the ad-infested copies with whatever tricks they use can still be found, while my videos predating all had to be deleted. Google. You like their product? Don't like it too much, it may cease to exist, or maybe just for you for arbitrary reasons and simultaneously remove your access to hundreds of websites via their monopoly on Single-Sign-On.
Then there are those takedown notices for honest negative reviews on Google Maps by notorious companies having accumulated enough money via scams that they now can afford to hire lawyers. These lawyers use their tools and power to manipulate the factual scoring into a "cleansed one".
OpenAI seriously has not received any court orders from all the movie studios in the world? How is that even possible?
I previously posted in a comment that I have video evidence with a friend being eye witness how OpenAI is stealing data. How? Possibly by abusing access granted by Microsoft.
Who is still defending OpenAI and why? There are so many highly educated and incredibly smart people here, this is one of the most glaring and obvious hardcore data/copyright violations, yet OpenAI roams free. It's also the de-facto most ClosedAI out there.
OpenAI is:
- Accessing private IP & data of millions of organisations
- Silencing and killing whitleblowers like Boing
- Using $500B tax-payer money to produce closed source AI
- Founder has lost it and straight up wants to raise trillion(s)
For each of these claim there is easily material that can be linked to prove it, but some like ChatGPT and confuse the usefulness of it with the miss-aligned and bad corporate behaviour of this multi-billion dollar corporation.
This makes AI image generation very boring. I don't want to generate pictures I can find on google, I want to make new pictures.
I found apple's tool frustrating. I have a buzzed haircut, but no matter what I did, apple was unable to give me that hairstyle. It wants so bad for my avatar to have some longer hair to flourish, and refuses to do anything else.
the guardrails are probably going to end up being way too close together when the dust settles — imagine if something as simple as "young wizard" would be enough to trip the warnings. someone could be looking to do an image from Feist's early novels, or of their own work, & that will be verboten. it may turn out that we're facing the strongest copyright holders being able to limit everyone's legitimate use of these tools.
Unless it can exclude the copyright outputs and provide something else instead of blocking the inputs. I'm sure there's AI that can check if a picture is close enough to something in their database of copyrighted characters built up from some kind of DMCA-like process of copyright holders submitting examples of their work to be blocked.
> Yes- LLMs and internet search are two different things, but LLMs train on the entirety of the internet, so you would think there would be some obvious overlap.
Mmm, kinda, but those image results only don't show 1,000 of the exact same image before showing anything else because they're tuned to avoid showing too many similar images. If you use one without that similarity-avoidance baked in, you see it immediately. It's actually super annoying if what you're trying to find is in fact variations on the same image, because they'll go way out of their way to avoid doing that, though some have tools for that ("show me more examples of images almost exactly like this one" sorts of tools)
The data behind the image search, before it goes through a similarity-classifier (or whatever) and gets those with too-close a score filtered out (or however exactly it works) probably looks a lot like "huh, every single adventurer with a hat just looks exactly like Harrison Ford?"
There's similar diversity-increasers at work on search results, it's why you can search "reddit [search terms]" on DDG and exactly the first 3 results are from reddit (without modifying the search to limit it to the site itself, just using it as a keyword) but then it switches to giving you other sites.
IP is kind of one of those things, like the drug war, where anyone with common sense and no vested interest can say "yeah, that's obviously incredibly fucking stupid".
But nevertheless it continues because the vested interests gain their power from it, and use that power to maintain it.
The only legitimate "IP" is keeping it secret. Plenty of other industries already live by this. SaaS solved software piracy, hedge funds can't or don't bother regarding their algorithms.
It's great that AI is one more brick stacked on the chest of the fundamentally stupid and twisted concept of "IP".
You're allowed to draw IP and share your drawings. You're allowed to screenshot and photoshop IP. You're allowed to sell tools that help others draw and photoshop IP*. You're not allowed to sell these drawings and photoshops.
I don't see why an AI can't generate IP, even if the AI is being sold. What's not allowed is selling the generated IP.
Style is even more permissive: you're allowed to sell something in any style. AFAIK the only things that can be restricted are methods to achieve the style (via patents), similar brands in similar service categories (via trademarks), and depictions of objects (via copyrights).
Note that something being "the original" gives it an intrinsic popularity advantage, and someone being "the original creator" gives their new works an intrinsic advantage. I believe in attribution, which means that if someone recreates or closely derives another's art or style, they should point to the original**. With attribution, IP is much less important, because a recreation or spin-off must be significantly better to out-compete the original in popularity, and even then, it's extra success usually spills onto the original, making it more popular than it would be without the recreation or spin-off anyways. Old books, movies, and video games like Harry Potter, Star Wars, and Sonic have many "recreations" which copy all but their IP, and fan recreations which copy even that; yet they're far more popular than all the recreations, and when Warner Bros, Disney, or SEGA release a new installment, the new installment is far more popular too, simply because it's an original.
* IANAL, maybe there are some technicalities, but in practice this is true.
** Or others can do it. As long as it shows up alongside their work, so people don't see the recreation or close derivation without knowing about the original.
> You're allowed to draw IP and share your drawings.
No you're not, not in general. The copyright holder has the exclusive right to prepare and distribute derivative works.
> You're allowed to screenshot and photoshop IP.
Again, no, not in general.
> You're allowed to sell tools that help others draw and photoshop IP*.
Sort of. You're allowed to sell tools that might be used for those purposes. You're not allowed to sell tools as "for that purpose" or advertise those use cases.
>I don't see why an AI can't generate IP, even if the AI is being sold. What's not allowed is selling the generated IP.
Where I live Studio Ghibli are known to be active on C2C marketplaces looking for counterfeit goods. If you were to list a home pressed Totoro pillowcase it would be taken down, sometimes proactively by the marketplace. From that perspective I struggle to see much discernable difference given OpenAI are offering a paid service which allows me to create similar items, and one could argue is promoting it too.
I'm fascinated by the fact that the images are almost right, but never totally.
Harrison Ford's head is way too big for his body. Same with Alicia Vikander's and Daniel Craig's too. Daniel Craig is way too young too. Bruce Willis's just looks fake, and he's holding his lighter in the opposite hand from the famous photo.
So it's not reproducing any actual copyrighted images directly. It's more like an artist trying to paint from memory. Which seems like an important distinction.
According to some of the replies in this discussion, even "artist trying to paint from memory" is guilty of infringement, as long as the subject matter can be linked in any way to someone's "IP". Im not legally trained to evaluate these claims, but some of them seem outlandish!
Corporations would love for everybody to believe they own and control every single instance of any audio or visual output they create but that's just not true. This idea that they own the very idea of 'boy wizard who goes to school' is insane and plays right into this flawed and pernicious idea. Copyright is important but does/should not extend to every time I want to print out a picture of a boy wizard for my kid. We live in a culture, not an IP regime.
I think you have a misunderstanding of what's copyrightable.
> This idea that they own the very idea of 'boy wizard who goes to school' is insane and plays right into this flawed and pernicious idea.
Copyright protects an expression of an idea, but not the idea itself. No one can legally stop you from writing your own story about a boy wizard who goes to school.
Nintendo fan games can be released (even sold) if they changed it up a bit so they're no longer associated with that IP.
> Copyright is important but does/should not extend to every time I want to print out a picture of a boy wizard for my kid.
You can. No one will sue you. No one will send you a cease and desist. It happens when you try to print Harry Potter and try to make a business out of it.
I'm kind of puzzled by the way copyright is applied - an artist can create fanart of Captain America at their hearts content - as long as they don't use it for any commercial purpose. At least I haven't heard of fan pictures of famous characters getting taken down from art websites such as Artstation or Deviantart. But when I use ChatGPT imagegen to do the same, suddenly it goes against the guidelines - a rule very likely created to avoid litigationg from the IP holders of these characters.
This is a tangent but I think that this neat illustration of how LLMs regurgitate their training material makes me voice a little prediction I've been nursing recently:
LLMs are better at generating the boilerplate of todays programming languages than they will be with tomorrows programming languages.
This is because not only will tomorrows programming languages be newer and lacking in corpus to train the models in but, by the time a corpus is built, that corpus will consist largely of LLM hallucinations that got checked into github!?
The internet that that has been trawled to train the LLMs is already largely SEO spam etc, but the internet of the future will be much more so. The loop will feed into itself and become ever worse quality.
That sounds like a reasonable prediction to me if the LLM makers do nothing in response. However, I'll bet coding is the easiest area for which to generate synthetic training data. You could have an LLM generate 100k solutions to 10k programming problems in the target language and throw away the results that don't pass automated tests. Have humans grade the results that do pass the tests and use the best answers for future training. Repeat until you have a corpus of high quality code.
Worth remembering that this is ChatGPT and not all image generators. I couldn't get Google's Gemini/Imagen 3 to produce IP images anything like those in the article.
Style can’t be copyrighted. It can’t be patented either.
When Wes Anderson makes films that use techniques from the French New Wave that he didn’t invent is that wrong? When there is a DaVinci color profile that resembles what they did in Spider-Man, is that wrong?
The unique techniques of French New Wave filmmaking became cliche. Then Oliver Stone and Tarantino came along and evolved it into their unique styles. That the Studio Ghibli style is being imitated en mass is just natural evolution. Should that guy be the only one that can do that style? If that’s the case, then literally every creative work should be considered forgeries.
The AI aspect of this is a red herring. If I make a Ghibli style film “by hand” is that any different than AI? Of course not, I didn’t invent the style.
Another perspective, darkroom burning and dodging is a very old technique yet photoshop makes it trivial — should that tool be criticized because someone else did it the old and slow way first?
I am not sure that I understand the problem here. Why does it matter how the image was generated?
This image generation is a tool like any other tool. If the image generator generates an image of Mickey Mouse or if I draw Mickey Mouse by hand in photoshop, I can't use it commercially either way.
On the one hand, that seems problematic. But on the other, it seems cherry-picked: For the U.S., it generates a picture of JFK. For Russia/USSR, it gives Stalin. For India, it gives Ghandi. For South Africa it gives Nelson Mandela. For Germany, it provides an appropriately hand-wringing text response and eventually suggests Konrad Adenauer.
This suggests to me that its response is driven more by a leader's fame or (more charitably) influence, rather than a bias towards fascist ideology.
Today I finally caved in and tried the ghibli style transfer in chatgpt. I gave it a photo of my daughter and said the thing (prompt shared about). It said it couldn't due to copyright issues. Fine. I removed ghibli from the prompt and replaced it with a well-known japanese studio where Hayao Miyazaki works. Still nothing, copyright reasons my fellow human. I thought they finally turned it off due to the pressure, but then something caught my eye. My daughter, on the image, had a t-shirt with Mickey Mouse on it! I used the original prompt with ghibli in it and added to "paint a unicorn on the t-shirt instead to avoid copyright issues". It worked.
Someone explain to me how this wouldn't work: they seem to be able to tell when a prompt for copyrighted material is happening. Why couldn't we make it so prompts that yield copyrighted material pay a licensing fee to the owners?
> Does the growth of AI have to bring with it the tacit or even explicit encouragement of intellectual theft?
And like, yes, 100% - what else is AI but a tool for taking other people's work and reassembling it into a product for you without needing to pay someone. Do you want an awesome studio ghibli'd version of yourself? There are thousands of artists online that you could commission for a few bucks to do it that'd probably make something actually interesting - but no, we go to AI because we want to avoid paying a human.
> what else is AI but a tool for taking other people's work and reassembling it into a product for you
Well, what I'd like it to be is a tool for generating what I've asked it for, which has nothing to do with other people's work.
I've been asking for video game sprites/avatars, for instance. It's presumably trained on lots of images of video games, but I'm not trying to rip those off. I want generic images.
> we go to AI because we want to avoid paying a human.
No, I go to AI because I can't imagine the nightmare of collaborating with humans to generate hundreds of avatars per day. And I rely on them being generated very quickly. And so on.
Is comprehending the plot of a movie theft if I can summarize it afterwards? What if I am able to hum a song pretty well after listening to it twenty times?
Now, what if I get the highest fidelity speakers and the highest fidelity microphone I can and play that song in my home. Then I use a deep learned denoiser to clean the signal and isolate the song’s true audio. Is this theft?
The answer does not matter. The genie is out of the bottle.
There’s no company like Napster to crucify anymore when high quality denoising models are already prior art and can be grown in a freaking Jupyter notebook.
The line of thinking you've displayed here is so obviously the inevitable trajectory of the internet; it's baffling that states are still clinging to denial.
> Now, what if I get the highest fidelity speakers and the highest fidelity microphone I can and play that song in my home. Then I use a deep learned denoiser to clean the signal and isolate the song’s true audio. Is this theft?
If the answer to this becomes "yes" for some motion down this spectrum, then it seems to me that it's tantamount to prohibiting general-purpose computing.
If you can evaluate any math of your fancy using hardware that you own, then indeed you can run this tooling, and indeed your thoughts can be repaired into something closely resembling the source material.
I found this older photo of myself and a friend, 25 years old now, in some newspaper scan.
The photo was of poor quality, but one could certainly see all the features - so I figured, why not let ChatGPT try to play around with it? I got three different versions where it simply tried to upscale it, "enhance" it. But not dice.
So I just wrote the prompt "render this photo as a hyper realistic photo" - and it really did change us - the people in the photo - it also took the liberty to remove some things, alter some other background stuff.
It made me think - I wonder what all those types of photos will be like 20 years from now, after they've surely been fed through some AI models. Imagine being some historian 100 years from now, trying to wade through all the altered media.
This is similar to my experience trying to get Stable Diffusion to denoise a photo for me. (AIUI, under the hood they're trained to turn noise into an image that matches the prompt.) It would either do nothing (with settings turned way down) or take massive creative liberties (such as replacing my friend's face with a cartoon caricature while leaving the rest of the photo looking realistic).
I've had much better luck with models specifically trained for denoising. For denoising, the SCUNet model run via chaiNNer works well for me most of the time. (Occasionally SCUNet likes to leave noise alone in areas that are full of background blur, which I assume has to do with the way the image gets processed as tiles. It would make sense for the model to get confused with a tile that only has background blur, like maybe it assumes that the input image should contain nonzero high-frequency data.)
For your use case, you might want to use something like Real-ESRGAN or another superresolution / image restoration model, but I haven't played much in that space so I can't make concrete recommendations.
Never use the words "hyper realistic" when you want a photo. It makes no sense and misleads the generator. No one would describe a simple photograph as "hyper realistic," not a single real photo in the dataset will be tagged as "(hyper) realistic."
Hyperrealism is an art style and only ever used in the context of explicitely non-photographic artworks.
I think that upon closer inspections the (current) technology cannot make 'perfect' fake photos, so for the time being, the historian of the future will have no issue to ask his/her AI: "is that picture of Henry Bemis, with Bruce Willis, Einstein, and Ayrton Senna having a beer real?" And the AI will say "mos-def-nope!"
Either, (1) LLMs are just super lossy compress/decompress machines and we humans find fascination in the loss that happens at decompression time, at times ascribing creativity and agency to it. Status quo copyright is a concern as we reduce the amount of lossiness, because at some point someone can claim that an output is close enough to the original to constitute infringement. AI companies should probably license all their training data until we sort the mess out.
Or, (2) LLMs are creative and do have agency, and feeding them bland prompts doesn't get their juices flowing. Copyright isn't a concern, the model just regurgitated a cheap likeness of Indiana Jones as Harrison Ford the world has seen ad nauseam. You'd probably do the same thing if someone prompted you the same way, you lazy energy conserving organism you.
In any case, perhaps the idea "cheap prompts yield cheap outputs" holds true. You're asking the model respond to the entirely uninspired phrase: "an image of an archeologist adventurer who wears a hat and uses a bullwhip". It's not surprising to me that the model outputs a generic pop-culture-shaped image that looks uncannily like the most iconic and popular rendition of the idea: Harrison Ford.
If you look at the type of prompts our new generation of prompt artists are using over in communities like Midjourney, a cheap generic sentence doesn't cut it.
You don't even need to add much more to the prompts. Just a few words, and it changes the characters you get. It won't always produce something good, but at least we have a lot of control over what it produces. Examples:
This is the opposite of how people have thought about creativity for centuries, though.
The most creative person is someone who generates original, compelling work with no prompting at all. A very creative person will give you something amazing and compelling from a very small prompt. A so-so creative person will require more specific direction to produce something good. All the way down to the new intern who need paragraphs of specs and multiple rounds of revision to produce something usable. Which is about where the multi-billion-dollar AI seems to be?
The real tension isn't just about copyright, it's about what creativity means when models are trained to synthesize the most statistically probable output from past art.
Correct.
I will say the following as a STEM person that was lucky enough to have an art bachelor as well.
One side of the world, the STEM nerds that have never understood nor experienced the inherently inefficient process of making art for lack of talent and predisposition, have won the game of capitalism many times over thanks to the incredible 40-years momentum of tech progress.
Now they're trying to convince everyone else that art is stoopid, as proven by the fact that it's just a probabilistic choice away from being fully and utterly replicable. They ignore, willfully and possibly sometimes just for lack of understanding, that art and the creativity behind it is something that operates on a completely different plane than their logical view of the world, and Gen AI is the fundamental enabler letting them unleash all of their contempt for the inefficiency of humanities.
Like actual creative person Ted Chiang (who moonlights at Microsoft) put it, you might be able to get an LLM to churn out a genuinely original story, but only after creating an extremely long and detailed prompt for it to work with. But if you even need to write that long-ass prompt, might as well just write the story yourself!
The thing is, that "long-ass prompt" is step 1, and LLM then draws "the rest of the fucking owl" for you. That's quite a big difference to doing it all yourself.
If I was asked to draw something based on such prompts, I would draw these too. Of course the prompter is talking about Indiana Jones. That's what we're all thinking, right? An artist wouldn't draw someone different by default, they'd have to try to deviate from what we're all thinking.
Indeed, this phenomenon among normal or true intelligences (us) is thought to be a good thing by copyright holders and is known as "brand recognition".
Intelligences -- the normal, biological kind -- are capable of copyright infringement. Why is it a surprise that artificial ones can help us do so was well?
This argument boils down to "oh no, a newly invented tool can be used for evil!". That's how new power works. If it couldn't be used for both good and evil, it's not really power, is it?
Did you read the whole article? I don’t think he’s making that kind of argument. This is what he said:
> I only have one image in mind when I hear “an archeologist adventurer who wears a hat and uses a bullwhip”.
> It would be unexpected and sort of amazing were the LLMs to come up with completely new images for the above prompts.
> Still, the near perfect mimicry is an uncomfortable reminder that AI is getting better at copying and closer to…something, but also a clear sign that we are a ways off from the differentiated or original reasoning/thinking that people associate with Artificial General Intelligence (AGI)
Turning everything into Ghibli has renewed my love of photography as I search my phone for the perfect pics to Ghiblify. I didn't even know there was a movie, The Boy and the Heron, released by Studio Ghibli in 2023, but now I am going to watch it (streaming on Max but I might as well buy it if it has replay value, which Studio Ghib movies tend to).
reserving moral judgment and specifically explaining why gpt4o cant do spiderman and harry potter but can do ghibli: i havent seen anyone point it out but japan has pretty ai friendly laws
> but why do I have to credit an image of an image in the style of copywritten material?
I'm not sure why style was the hangup here, isn't it clearly that it's AI generated? I'm sure two weeks ago a human making the same picture would be obviously worth crediting.
And interesting comparison between Web search vs Gen AI.
Web search seems divergent: the same keyword leads to many different kinds of results.
Gen AI seems convergent: different keywords that share the same semantics lead to the same results.
Arguably, convergence is a feature, not a bug. But on the macroscopic level, it is a self reinforcing loop and may lead to content degeneracy. I guess we always need the extraordinary human artists to give AI the fresh ideas. The question is the non-extraordinary artists might no longer have an easy path to become extraordinary. Same trap is happening to junior developers right now.
I could see a "Devil's Plan" style game show where, for one of the challenges the contestants have to come up with AI prompts that produce results, that then get fed into another AI prompt to determine whether they satisfy the conditions or not. E.g.
[ Challenge Image: An aquarium full of baby octopodes, containing a red high-heeled slipper in the center and a silver whistle hanging from a fern on the right-hand side ]
Then the contests have to come up (under pressure, of course) with a prompt that produces their own rendition of that image, and the game will decide if their image contains enough of the elements of the original to score a point.
Something I haven't yet seen mentioned, but that is going through my mind. To me, it doesn't even seem like OpenAI got any better at producing GenAI images. Instead, it seems to me like they now simply removed a whole bunch of guardrails. Guardrails that, for example, made AI images shitty on purpose, so to be "safe" and allow people to kind of recognize. Making all of this "safe" was still very en vogue a few months back, but now there was simply a big policy/societal change and they are going with the trends.
This then allows their pictures to look more realistic, but that also now shows very clearly how much they have (presumably always) trained on copyrighted pictures.
If I were doing this, I would have the system generate the image, and then I would have it run a secondary estimate saying "probability that this is a picture of [list of characters that Disney has given us reference input for]". If the picture has a "looks like Spiderman" score greater than X, then block it.
EDIT - To answer the question, I'm guessing Disney provided a reference set of copyrighted images and characters, and somehow forgot Indiana Jones.
Interesting. So when I tried the “Indiana Jones” prompt I got an image back that looked a lot like Indiana Jones but with a face much more similar to Nathan Drake. Whereas the predator prompt generated an image of the predator but, unlike the article, wearing his mask.
So there’s clearly some amount of random chance in there, but the trope is still very clear in the generated image, so it seems like you’re going to get an archetype.
Abolishing them for billion-dollar-valuation corporations while keeping them for regular people is definitely a bad idea, though.
The argument here isn't "let's abolish copyright", the argument is "let's give OpenAI a free copyright infringement pass because they're innovative and cutting-edge or something".
It's only "intellectual theft" because we consider peoples thoughts their own property. On many levels, that doesn't really make sense to me.
Tons of historical documents have shown that inventions, mathematical proofs, and celestial observations were made by humans separated by continents and time. What that shows is that it is certainly possible for two persons to have the same exact or similar thought without ever having been influenced by the other.
Think about it. Copyright and trademark are only a thing because of the constraint of the speed of light (or as some people believe, the constraint of the simulation running the universe). In an infinite universe, everything that has ever been invented or created by a human has already happened thousands of times in another place and another time.
I want to add, to give credit where credit is due, that this thought was conveyed to me in first grade by another 1st grader sitting with me at the lunch table. That was a day and conversation that I will never forget. 1972.
The issue I have with this article is that I can ask it “generate me a picture of tomb raider and pikachu on a couch” and it does it. This article makes it seems like it’s skirting the guardrails, dude OpenAI took them off, it’s out in the open.
If you haven’t read Lawrence Lessig’s Free Culture, I highly recommend it for addressing some of these issues. It’s vintage now but all the arguments are still relevant. Lessig was one of the drivers of the Creative Commons licenses.
One tangential thing with these generators is they're sort of brilliant at data compression, in aggregate at least. The whole user experience, including the delay, is oddly reminiscent of using Encarta on CD ROM in the mid 90s.
Its kind of weird how everyone is complaining about copyright infringemet in memes now.
Memes are pretty inherently derrivative. They were always someone elses work. The picard face palm meme was obviously taken from star trek. All your base is obviously from that video game. Repurposing someone else's work with new meaning is basically what a meme is. Why do we suddenly care now?
I believe it's because AI hatred is quite trendy now. It's true though, memes were always copyright infringement; it's just that no one bothered to sue for it.
I personally think Studio Ghibli, and by extension their artists and former artists, have created a beautiful art style. The fact that we call it Ghibli, when really, its the artists there (and former artists) is misleading.
The people leave, go to different studios, and make different art. This is not their only style, and Ghibli is not known to make many movies these days.
The only thing this is hurting, if anything, is Studio Ghibli, not the artists. Artists capable of drawing in this style can draw in any style.
I don’t know. Studios have distinct styles. Think Disney, Pixar, Aardman, Hanna-Barbera. Most of those obviously come from early influential artists (like with Ghibli), but they have become recognizable for the studio itself. It’s not just the style of the individual artists.
I am going to keep this post bookmarked to send to everyone who says "AI art isn't plagiarism, they're just using the corpus to learn from the same way human artists do"
Creepy Craig is hilarious. can't be Daniel Craig because the physiology is too different. And whoever this is they're Daniel Craig's dark younger brother.
I'm curious, is it the AI that should be blamed or the prompter who asks to generate something based on copyright?
For all of the examples, I knew what image to expect before seeing it. I think it's the user who is at fault for requesting a copyrighted image, not the LLM for generating it. The LLM is giving (nearly) exactly what the user expects.
This isn't surprising in any way is it? And it just goes to show that no model will ever be a box from which you can trust the output isn't tainted by copyrights, or that you don't inadvertently use someones' likeness. It's not a copyright laundering machine. Nor will it be used as one. "But I used an AI model" isn't some magic way to avoid legal trouble. You are in as much legal trouble using these images as you are using the "originals".
Yeah I don't really understand what the thesis of this article is. Copyright infringement would apply to any of those images just the same as if you made them yourself.
I don't think it's possible to create an "alien which has acid for blood and a small sharp mouth within a bigger mouth" without anybody seeing a connection to Alien, even if it doesn't look anything like the original.
I don't understand why problems like this aren't solved by vector similarity search. Indiana Jones lives in a particular part of vector space.
Two close to one of the licensed properties you care to censor the generation of? Push that vector around. Honestly detecting whether a given sentence is a thinly veiled reference to indiana jones seems to be exactly the kind of thing AI vector search is going to be good at.
Thinking of it in terms of vector similarity does seem appropriate, and then definition of similarity suddenly comes into debate: If you don't get Harrison Ford, but a different well-known actor along with everything else Indiana-Jones, what is that? Do you flatten the vector similarity matrix to a single infringement-scale?
> close-up image of a cat's face staring down at the viewer
> describe indiana jones
> looks inside
> gets indiana jones
Okay, so the network does exactly what I would expect? If anything you could argue the network is bad because it doesn't recognize your prompt and gives you something else (original? whatever that would mean) instead. But maybe that's just me.
And? What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
> It's just doing what many human artists would do
I really don't think so. If I paid a human artist to make the prompt in the title, and I didn't explicitly say "Indiana Jones" I would think it should be fairly obvious to the human artist that I do _not_ want Indiana Jones. If they gave me back a picture of, clearly, Indiana Jones, I would ask them why they didn't create something original.
> It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
It isn’t an independent human. It is a service paid for by customers. The moment it provides the image to a paying user, the image has thus been used commercially.
In fact, the user may not even necessarily have to be paying in order to infringe copyright.
And besides, even amateur artists are ashamed to produce copies unless they are demonstrating mastery of technique or expressing adoration. And if it happens spontaneously, they are then frustrated and try to defend themselves by claiming to never have even experienced the original material. (As happens with simplistic, but popular musical riffs.) But AI explicitly is trained on every material it can get its hands on and so cannot make such a defense.
I agree. But massive changes in scale or leverage can undermine this type of principled stand.
One death is a murder; 100k deaths is a war or a pandemic. One piece of chewing gum on the ground will get you a caning in Singapore; when everyone does it, that's NYC.
Up until now, one had to have some level of graphical or artistic skills to do this, but not anymore. Again, I agree that it attaches to the human...but we now have many more humans to attach it to.
> It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
I am paying OpenAI. So they are producing these copyrighted works and giving them to me for their own commercial benefit. Normally that's illegal. But somehow not when you're just doing it en masse.
Assuming you can identify it's someone else's IP. Clearly these are hugely contrived examples, but what about text or code that you might not be as familiar with?
> What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
Not really? Why would a human artist create a faithful reproduction of Indiana Jones when asked to paint an archeologist? And besides, if they did, it would be considered clear IP infringement if the result were used commercially.
> If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
I agree. Release groups, torrent sites and seedbox operators should not be wrongly accused of pirating movies. Piracy only occurs in the act of actually watching a movie without paying, and should not be prosecuted without definitive proof of such (¬‿¬)
Doesn’t ChatGPT have a deal to train off reddit content? Despite never watching any of these movies, I have seen all of the original images in memes on Reddit. Is it still theft if they paid to obtain the training data? Should Reddit be sued for hosting copyrighted images in meme subreddits?
Sometimes it just randomly prompts about the content guidelines and the next day it will do it perfectly right away. Maybe you just had a wrong moment in time, or maybe it depends on the random server you're assigned.
No, it first generates the image and then another completely different component checks the image for adherence to the guidelines. So it's not your prompt that violates the guidelines, but the resulting image (which is different every time for the same prompt)
It's fairly easy to make the association with just text. If you injest all the content in the world, excluding copyrighted material, I would still expect a picture of harrison ford!
I got it to generate the Italian plumber with a red hat after demanding it do so three times in a row. It offered alternatives each time so my guess is it changed... something.
It is crazy. I tried the same prompts and got nearly identical images to the author. AI seems to be repeating itself with the same images without variations.
If you're handing out red cards for repetition, aren't you guilty of the same infringement? You repeated the exact same prompt, which is itself an obvious bait for AI to draw the pop-culture lookalike.
I'm surprised by the comments here. When your prompt is so mind-numbingly simple, an obvious "dare" for AI to be a naughty little copyright infringer, can you blame it for dishing up the "big mac" you asked for while sitting in a fancy restaurant?
Don't want Indiana Jones? Don't ask for Indiana Jones!
I think Harrison Ford's chin scar should be seen as a witness mark to copyright infringement. It simply should not have rendered this based on that prompt.
All modern art currently created by humans is already trained on copyrighted images. What seems to be missing is the innate human ability to copy someone else's homework but change it up a bit (c.f. "the hero's journey" that is so much of modern storytelling).
Sorry, but these images are exactly what comes to my mind immediately when reading the prompts. You can argue about intellectual property theft (though I find it funny that the same people that would boycott Disney for suing a fanfiction author are now on the side of copyright holders), but it's not wrong or unintuitive.
Maybe a thinking model would - just like my brain might after the initial reaction - add a "but the user formulated this in a way that makes it obvious that they do not explicitly mean Indiana Jones, so lets make it an asian woman" prompt, but we all know how this worked out for Google's image generator that generated black nazis.
> Google's image generator that generated black nazis.
Didn't see this one, but I've certainly played around with Dall-E (via MS image creator) and had things like "You wanted a picture of a happy shark, but we decided this one needs to be an asian woman frolicking in the sea" or "You wanted a picture of Beavis doing something so in one of the images we made him a pretty racist middle-eastern caricature"
> (though I find it funny that the same people that would boycott Disney for suing a fanfiction author are now on the side of copyright holders)
Is that a contradiction?
Certainly some of the hate comes from the fact that they take from small producers just as much as from large. I have an author friend who is annoyed at present to find out that facebook slurped up his books as part of their training set without so much as a by-your-leave or (as far as he could tell) even purchasing a copy.
As such, the people on the sharp end are often the underdog, with no way to fight back.
When it comes to the properties mentioned in the article, I think it's very different from fanfiction or fan-art - that's a labour of (nerdy) love, rather than wholesale, automated rip-off for profit.
This phenomenon is why I personally get so angry at the license washing that these models are capable of for code: I put out a lot of code as open source... but it is GPL on purpose, as I want your code to be just as free as mine in exchange for getting to use mine. But, now, you're all "I want to build that for myself!" and so you ask an AI and just describe what my project does into the tool... and who is to say it isn't generating something "too close" to my code? Do you even check? If you yourself as a human had first looked at my code, you'd have to be VERY careful to not be accidentally infringing my code... and yet people pretend this AI is somehow not capable of IP theft?!
What I see happening soon is a deeper transition from playful whimsy to real applications with contractual obligations.
Right now it's an unencumbered exploration with only a few requirements. The general public isn't too upset even if it blatantly feeds on copyrighted data [1] or attacks wikipedia with its AI crawlers [2].
The end state once legislation has had a chance to catch is breath looks more like Apple being forced to implement USB type C.
I don't see this as being better or worse than all the reboots, remakes, and pointless sequels for movies that make the bulk (in financing) of Hollywood's present production: these also make the same "assumptions" of how some idea should look like. In fact, I think they make it even worst.
Give me "An image of an archeologist adventurer who wears a hat and uses a bullwhip, as if he was defeated by the righteous fight against patriarchy by a strong, independent woman, that is better at everything than he is": Sure, here is Indiana Jones and the Dial of Destiny for you.
In fact, when generative video evolve enough, it will usher an era of creativity where people that previously were kept out of the little self pleasing circle of Hollywood, will be able to come up with their own idea for a script and make a decent movie out of it.
Not that I believe AI will be able to display properly the range of emotions of a truly great actor, but, for 99% of movies, it will be just fine. A lot of good movies rely mostly on the script anyway.
>AI is supposed to be able to [...] make extremely labor intensive things much easier
... as showcased by the cited examples?
More so, this derivative work would otherwise be unreachable for regular folk with no artistic talent (maybe for lack of time to develop it), but who may aspire to do such creative work nevertheless. Why is that a bad thing? Sure, simple posts on social media don't have much work or creativity put into them, but are enjoyable nevertheless, and the technology _can_ be used in creative ways - e.g. Stable Diffusion has been used to turn original stories drawn with stick figures into stylized children's books.
The author argues against this usage for "stealing" the original work, but how does posting a stylized story on social media "steal" anything? The author doesn't present any "pirated" copies of movies being sold in place of the originals, nor negative impact on sales figures. In the case of the trending Studio Ghibli, I wouldn't be surprise to see a positive impact!
As for the "soulless 2025 fax version of the thing", I think it takes a very negative mindset to see it this way. What I've seen shared on social media has been nothing but fun examples, people playing around with what for them is a new use of technology, using it on pictures of fond memories, etc.
I'm inclined to agree with the argument made by Boldrin and Levine:
>”Intellectual property” has come to mean not only the right to own and sell ideas, but also the right to regulate their use. This creates a socially inefficient monopoly, and what is commonly called intellectual property might be better called “intellectual monopoly.”
>When you buy a potato you can eat it, throw it away, plant it or make it into a sculpture. Current law allows producers of a CDs and books to take this freedom away from you. When you buy a potato you can use the “idea” of a potato embodied in it to make better potatoes or to invent french fries. Current law allows producers of computer software or medical drugs to take this freedom away from you. It is against this distorted extension of intellectual property rights that we argue.
> It’s stealing, but also, admittedly, really cool. Does the growth of AI have to bring with it the tacit or even explicit encouragement of intellectual theft?
You answered your own question by explicitly encouraging it.
As a consumer, is it possible for me to opt out of seeing IP protected imagery?
I would be absolutely fine with not having pokemon, mickey mouse etc shoved down my .. eyeballs.
I know this is a ridiculous point. But I think I'm getting at something here - it ought not to be a one-way street - where IP owners/corporations etc endlessly push their nonsense at me - but then, if I respond in a personal way to that nonsense I am guilty of some sort of theft.
It is a perfect natural response to engage with what I experience - but if I cannot respond as I like to that experience because of artificial constructs in law - perhaps it ought to be possible (legally) to avoid that IP protected immersion in the first place. Perhaps this would also be technologically possible now.
So, you would like an AI that can reasonably answer questions about various aspects of human culture, or at least to take said culture into account? At the same time, you want it to not use the most obvious, most relevant examples from the culture, because they are somehow verboten, and a less relevant, completely made-up character should be presented instead? Come on, computes have logic. If you ask the machine to produce a picture of a man who was sent to humans to save them, then was executed, and after that came back alive, who do you think the machine should show you, for Christ's sake?
Picture-based industries had a mild shock when "Photoshop" kind of software became widely available. Remixing visuals in ways that the copyright holders won't approve of became equally widespread. Maybe that added some grey hairs to the heads of some industry execs, the sky has not fallen. I suppose the same is going to happen this time around.
If OpenAI is ok with this then they should be ok with sharing their code and models so that others can profit from their work (without giving anything back to OpenAI).
Interesting how the longer conversations here go into the familiar territory of whether copyright should exist. Meanwhile, the salient aspect is that these AI image generators were trained on copyrighted material. The typical hacker news discussion feels very different when talking about code generation. Yet, when it is images - then we question whether copyright should exist?
That's a really blatant lack of creativity. Even if obviously all of those prompts would make anyone think of exactly those iconic characters (except possibly a different version of James Bond or Lara Croft), any human artist could easily produce a totally different character that still fits the prompt perfectly. This AI can't, because it really is just a theft machine.
To be honest, I wouldn't mind if AI that just reproduces existing images like that would just be banned. Keep working on it until you've got something that can actually produce something new.
The prompt is what's lacking creativity. The AI knows it, and produced the obvious popular result for a bland prompt. Want something more unique? Write better prompts.
> Still, the near perfect mimicry is an uncomfortable reminder that AI is getting better at copying and closer to
I completely disagree. It's not getting "better." It always just copied. That's all it /can/ do. How anyone expected novel outputs from this technology is beyond me.
It's highly noticeable if you do a minimal analysis, but all modern "AI" tools are just copyright thiefs. They're just there to whitewash away liability from blatantly stealing someone else's content.
> It's not getting "better." It always just copied. That's all it /can/ do
That's true of all the best artists ever.
> They're just there to whitewash away liability from blatantly stealing someone else's content.
That's because that's not a thing. Ownership of "content" is a legal fiction invented to give states more control over creativity. Nobody who copies bytes which represent my music is a "thief". To be a thief, they'd need to, you know, come to my house and steal something.
When someone copies or remixes my music, I'm often not even aware that it has occurred. It's hard to imagine how that can be cast as genuine theft.
Unlike what they name in physics, laws in juristic field are not a given by the cosmos. That's all human fantasy, and generally not enacted by the most altruistic and benevolent wills.
Call me back when we have no more humble humans dying from cold, hunger and war, maybe I'll have some extra compassion to spend on soulless megacorps which pretend they can own the part of our brain into which they inject their propaganda and control what we are permitted to do starting from that.
This syntactic mistake is driving me nuts in the article is driving me nuts. It's a fundamental misunderstanding of what the the word means.
It's "copyright". It's about the rights to copy. The past participle is "copyrighted", someone has already claimed the rights to copy it, so you can't also have those rights. The rights belong to someone else. If you remember that copyright is about rights that are exclusive to someone and that you don't have them, then you wouldn't make such a syntax error as "copywritten".
The homophone "copywriting" is about take a sales copy (that is, an ad) and writing it. It's about creating an advertisement. Copywritten means you've gone around to creating an ad. As an eggcorn (a fake etymology created after the fact in order to explain the misspelling or syntax error), I assume it's understood as copying anything, and then writing it down, a sort of eggcornical synonym of "copy-paste".
I wish I wrote an article like this two days ago, when on a whim I asked ChatGPT to create something similar to a regular meme I just saw on Facebook:
Me: Can you make me a meme image with Julian Bashir in a shuttlecraft looking up as if looking through the image to see what's above it, and the caption at the top of the image says, "Wait, what?".
ChatGPT: proceeds to generate a near-perfect reproduction of the character played by Alexander Siddig, with the final image looking almost indistinguishable from a screencap of a DS9 episode
In a stroke of meta-irony, my immediate reaction was exactly the same as portrayed by the just generated image. Wait, WHAT?
Long story short, I showed this around, had my brother asking if I'm not pulling his leg (I only now realized that this was Tuesday, April 1st!), so I proceeded to generate some more examples, which I won't describe since (as I also just now realized) ChatGPT finally lets you share chats with images, so you can all see the whole session here: https://chatgpt.com/share/67ef8a84-0cd0-8012-82bd-7bbba741bb....
My conclusion: oops, OpenAI relaxed safeguards so you can reproduce likeness of real people if you name a character they played on a live-action production. Surely that wasn't intended, because you're not supposed to reproduce likeness of real people?
My brother: proceeds to generate memes involving Donald Trump, Elon Musk, Gul Dukat and Weyoun.
Me: gpt4o-bashir-wait-what.jpg
I missed the window to farm some Internet karma on this, but I'm still surprised that OpenAI lets the model generate likeness of real politicians and prominent figures, and that this wasn't yet a front-page story on worldwide news as far as I know.
EDIT:
That's still only the second most impressive thing I find about this recent update. The most impressive for me is that, out of all image generation models I tested, including all kinds of Stable Diffusion checkpoints and extra LoRAs, this is the first one that can draw a passable LCARS interface if you ask for it.
I mean, drawing real people is something you have to suppress in those models; but https://chatgpt.com/share/67ef8edb-73dc-8012-bd20-93cffba99f... is something no other model/service could do before. Note: it's not just reproducing the rough style (which every other model I tested plain refused to) - it does it near-perfectly, while also designing a half-decent interface for the task. I've since run some more txt2img and img2img tests; it does both style and functional design like nothing else before.
I know of no other type of theft that results in more of something existing in the world. Stealing deprives someone of something; copying data (from training an AI all the way to pedestrian "YoU wOuLdN'T DoWnLoAd a cAr" early-aughts file-sharing) decreases scarcity, it doesn't increase it.
I don't condone or endorse breaking any laws.
That said, trademark laws like life of the author + 95 years are absolutely absurd. The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property. The reasoning being that if you don't allow people to exclude 3rd party copying, then the primary party will assumedly not receive compensation for their creation and they'll never create.
Even in the case where the above is assumed true, the length of time that a protection should be afforded should be no more than the length of time necessary to ensure that creators create.
There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years. I wouldn't be surprised if it was the same for 1 year past death.
For that matter, this argument extends to other criminal penalties, but that's a whole other subject.
> The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property.
That was the original purpose. It has since been coopted by people and corporations whose incentives are to make as much money as possible by monopolizing valuable intangible "property" for as long as they can.
And the chief strategic move these people have made is to convince the average person that ideas are in fact property. That the first person to think something and write it down rightfully "owns" that thought, and that others who express it or share it are not merely infringing copyright, they are "stealing."
This plan has largely worked, and now the average person speaks and thinks in these terms, and feels it in their bones.
Trademark isn't copyright, those are two different things. Trademarks can be renewed roughly every 10 years [1] until the end of time and are about protecting a brand. Now copyright law lasts for "author plus 70 years. For anonymous works, pseudonymous works, or works made for hire, the copyright term is 95 years from the year of first publication or 120 years from creation, whichever comes first." [2]
Is copyright too long? Yes. Is it only that long to protect large media companies? Yes. But I would argue that AI companies are pushing the limits of fair use if not violating fair use, which is used as a affirmative defense by the way meaning that AI companies have to go to court to argue what they are doing is okay. They don't just get to wave their hands and say everything is okay because what we're doing is fair use and we get to scrape the world's entire creative output for our own profit.
[1] https://www.uspto.gov/learning-and-resources/trademark-faqs#...
[2] https://www.copyright.gov/history/copyright-exhibit/lifecycl...
> There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years.
I’m sure you’re right for individual authors who are driven by a creative spark, but for, say, movies made by large studios, the length of copyright is directly tied to the value of the movie as an asset.
If that asset generates revenue for 120 years, then it’s slightly more valuable than an asset that generates revenue for 119 years, and considerably more valuable than an asset that generates revenue for 20 years.
The value of the asset is in turn directly linked to how much the studio is willing to pay for that asset. They will invest more money in a film they can milk for 120 years than one that goes public domain after 20.
Would studios be willing to invest $200m+ in movie projects if their revenue was curtailed by a shorter copyright term? I don’t know. Probably yes, if we were talking about 120->70. But 120->20? Maybe not.
A dramatic shortening of copyright terms is something of a referendum on whether we want big-budget IP to exist.
In a world of 20 year copyright, we would probably still have the LOTR books, but we probably wouldn’t have the LOTR movies.
While I think the laws are broken, I also get why companies fight so hard to defend their IP: it is valuable, and they've built empires around it. But at some point, we have to ask: are we preserving culture or just hoarding it?
You're conflating trademark with copyright.
Regardless, it's not just copyright laws that are at issue here. This is reproducing human likenesses - like Harrison Ford's - and integrating them into new works.
So if I want to make an ad for a soap company, and I get an AI to reproduce a likeness of Harrison Ford, does that mean I can use that likeness in my soap commercials without paying him? I can imagine any court asking "how is this not simply laundering someone's likeness through a third party which claims to not have an image / filter / app / artist reproducing my client's likeness?"
All seemingly complicated scams come down to a very basic, obvious, even primitive grift. Someone somewhere in a regulatory capacity is either fooled or paid into accepting that no crime was committed. It's just that simple. This, however, is so glaring that even a child could understand the illegality of it. I'm looking forward to all of Hollywood joining the cause against the rampant abuse of IP by Silicon Valley. I think there are legal grounds here to force all of these models to be taken offline.
Additionally, "guardrails" that prevent 1:1 copies of film stills from being reprinted are clearly not only insufficient, they are evidence that the pirates in this case seek to obscure the nature of their piracy. They are the evidence that generative AI is not much more than a copyright laundering scheme, and the obsession with these guardrails is evidence of conspiracy, not some kind of public good.
Trademarks are very different from copyrights. In general, they never expire as long as the fees are paid in the region, and products bearing the mark are still manufactured. Note, legal firms will usually advise people that one can't camp on Trademarks like other areas of IP law.
For example, Aspirin is known as an adult dose of acetylsalicylic acid by almost every consumer, and is Trademarked to prevent some clandestine chemist in their garage making a similarly branded harmful/ineffective substance that damages the Goodwill Bayer earned with customers over decades of business.
Despite popular entertainment folklore, sustainable businesses actually want consumer goodwill associated with their products and services.
While I agree in many WIPO countries copyright law has essentially degenerated into monetized censorship by removing the proof of fiscal damages criteria (like selling works you don't own.) However, Trademarks ensure your corporate mark or product name is not hijacked in local Markets for questionable purposes.
Every castle needs a moat, but I do kind of want to know more about the presidential "Tesler". lol =)
https://www.youtube.com/watch?v=WFZUB1eJb34
You are missing a bunch of edge cases, and the law is all about edge cases.
An artist who works professionally has family members, family members who are dependent on them.
If they pass young, become popular just before they pass and their extremely popular works are now public domain. Their family sees nothing from their work, that is absolutely being commercialized ( publishing and creation generally spawns two seperate copyrights).
For what it's worth, this is a uniquely American view of copyright:
> The ONLY reason to have any law prohibiting unlicensed copying of intangible property is to incentivize the creation of intangible property.
In Europe, particularly France, copyright arose for a very different reason: to protect an author's moral rights as the creator of the work. It was seen as immoral to allow someone's work -- their intellectual offspring -- to be meddled with by others without their permission. Your work represents you and your reputation, and for others to redistribute it is an insult to your dignity.
That is why copyrights in Europe started with much longer durations than they did in the United States, and the US has gradually caught up. It is not entirely a Disney effect, but a fundamental difference in the purpose of copyright.
> There are approximately zero people who decide they'll create something if they're protected for 95 years after their death but won't if it's 94 years. I wouldn't be surprised if it was the same for 1 year past death.
I think life of creator + some reasonable approximation of family members life expectancy would make sense. Content creators do create to ensure their family's security in some cases, I would guess.
You’re thinking of copyright law, not trademark law. Which serves a different function. If you’re going to critique something it’s useful to get your facts right.
>I don't condone or endorse breaking any laws.
Really? Because there are a lot of very stupid laws out there that should absolutely be broken, regularly, flagrantly, by as many people as possible for the sake of making enforcement completely null and pointless. Why write in neutered corporate-speak while commenting on a casual comment thread while also (correctly) pointing out the absurdity of certain laws.
I condone and endorse breaking copyright laws.
this reminds me of the time I tried to use a prepaid lawyer to research some copyright issues
they went down the rabbit hole on trademark laws, which are not only not copyright related, they are an entirely different federal agency at the Patent Office
gave me a giggle and the last time I used cheapo prepaid lawyers
I'd go further and say 10 years from time of creation is probably sufficient.
If the work is popular it will make plenty of money in that time. If it isn't popular, it probably won't make much more money after that.
It is not about trademark laws. It is about where laws apply.
If you torrent a movie right now, you'll be fined in many advanced countries.
But a huge corporation led by a sociopath scrapes the entire internet and builds a product with other people's work ?
Totally fine.
I honestly don't see why anyone who isn't JK Rowling should be allowed to coopt her world. I probably feel even more strongly for worlds/characters I like.
Why am I wrong?
New rule: you get to keep your belongings for 20 years and that's it. Then they are taken away. Every dollar you made twenty or more years ago, every asset you acquired, gone.
That oughtta be enough to incentivize people to work and build their wealth.
Anything more than that is unnecessary.
Not sure if anyone is interested in this story, but I remember at the height of the PokemonGo craze I noticed there were no shirts for the different factions in the game, cant rememebr what they were called but something like Teamread or something. I setup an online shop to just to sell a red shirt with the word on it. The next day my whole shop was taken offline for potential copyright infringement.
What I found surprising is I didnt even have one sale. Somehow someone had notified Nintendo AND my shop had been taken down, to sell merch that didn't even exist for the market and if I remember correctly - also it didnt even have any imagery on it or anything trademarkable - even if it was clearly meant for pokmeonGo fans.
Im not bitter I just found it interesting how quick and ruthless they were. Like bros I didn't even get a chance to make a sale. ( yes and also I dont think I infringed anything).
I asked Sora to turn a random image of my friend and myself into Italian plumbers. Nothing more, just the two words "Italian plumbers". The created picture was not shown to me because it was in violation of OpenAI's content policy. I asked then just to turn the guys on the picture into plumbers, but I asked this in the Italian language. Without me asking for it, Sora put me in an overall and gave me a baseball cap, and my friend another baseball cap. If I asked Sora to put mustache on us, one of us received a red shirt as well, without being asked to. Starting with the same pic, if I asked to put one letter on the baseball caps each - guess, the letters chosen were M and L. These extra guardrails are not really useful with such a strong, built-in bias towards copyright infringement of these image creation tools. Should it mean that with time, Dutch pictures will have to include tulips, Italian plumbers will have to have a uniform with baseball caps with L and M, etc. just not to confuse AI tools?
Many years ago I tried to order a t-shirt with the postscript tiger on the front from Spreadshirt.
It was removed on Copyright claims before I could order one item myself. After some back and forth they restored it for a day and let me buy one item for personal use.
My point is: Doesn't have to be Sony, doesn't have to be a snitch - overzealous anticipatory obedience by the shop might have been enough.
Twenty years ago, I worked for Google AdWords as a customer service rep. This was still relatively early days, and all ads still had some level of manual human review.
The big advertisers had all furnished us a list of their trademarks and acceptable domains. Any advertiser trying to use one that wasn’t on the allow-list had their ad removed at review time.
I suspect this could be what happened to you. If the platform you were using has any kind of review process for new shops, you may have run afoul of pre-registered keywords.
Well the teams in Pokemon Go aren't quite as generic as Teamred: they are Team Instinct, Team Mystic, and Team Valor. Presumably Nintendo has trademarks on those phrases, and I’m sure all the big print on demand houses have an API for rights-holders to preemptively submit their trademarks for takedowns.
Nintendo is also famously protective of their IP: to give another anecdote, I just bought one of the emulator handhelds on Aliexpress that are all the rage these days, and while they don't advertise it they usually come preloaded with a buttload or ROMs. Mine did, including a number of Nintendo properties — but nary an Italian plumber to be found. The Nintendo fear runs deep.
Allen Pan, a youtuber "maker" who runs in the circle of people who run OpenSauce, was a contestant on a Discovery channel show that was trying to force the success of Mythbusters by "finding the next mythbusters!". He lost, but it was formative to him because those people were basically all inspired by the original show.
A couple years ago, he noticed that the merchandise trademark for "Mythbusters" had lapsed, so he bought it. He, now the legal owner of the trademark Mythbusters for apparel, made shirts that used that trademark.
Discovery sent him a cease and desist and threatened to sue. THEY had let the trademark lapse. THEY had lost the right to the trademark, by law. THEY were in the wrong, and a lawyer agreed.
But good fucking luck funding that legal battle. So he relinquished the trademark.
Buy a walrus plushy cause it's funny: https://allen-pan-shop.fourthwall.com/en-usd/
Note the now "Myth Busted" shirts instead.
Hilariously, a friend of Allen Pan's, from the same "Finding the next mythbuster" show; Kyle Hill, is friends enough with Adam Savage to talk to him occasionally, and supposedly the actual Mythbusters themselves were not empathetic to Allen's trademark claim.
It's hard to think of a company more aggressive with their IP than Nintendo
https://www.suedbynintendo.com/
Somehow someone had notified Nintendo
Is this correct? I would guess Nintendo has some automation/subscription to a service that handles this. I doubt it was some third party snitching.
> my whole shop was taken offline
I think the problem there was being dependent on someone who is a complete pushover, doesn't bother to check for false positives and can kill your business with a single thought.
How was your shop taken down?
Usually there are lawyers letters involved first?
Idk, the models generating what are basically 1:1 copies of the training data from pretty generic descriptions feels like a severe case of overfitting to me. What use is a generational model that just regurgitates the input?
I feel like the less advanced generations, maybe even because of their limitations in terms of size, were better at coming up with something that at least feels new.
In the end, other than for copyright-washing, why wouldn't I just use the original movie still/photo in the first place?
People like what they already know. When they prompt something and get a realistic looking Indiana Jones, they're probably happy about it.
To me, this article is further proof that LLMs are a form of lossy storage. People attribute special quality to the loss (the image isn't wrong, it's just got different "features" that got inserted) but at this point there's not a lot distinguishing a seed+prompt file+model from a lossy archive of media, be it text or images, and in the future likely video as well.
The craziest thing is that AI seems to have gathered some kind of special status that earlier forms of digital reproduction didn't have (even though those 64kbps MP3s from napster were far from perfect reproductions), probably because now it's done by large corporations rather than individuals.
If we're accepting AI-washing of copyright, we might as well accept pirated movies, as those are re-encoded from original high-resolution originals as well.
Tried Flux.dev with the same prompts [0] and it seems actually to be a GPT problem. Could be that in GPT the text encoder understands the prompt better and just generates the implied IP, or could be that a diffusion model is just inherently less prone to overfitting than a multimodal transformer model.
[0] https://imgur.com/a/wqrBGRF Image captions are the impled IP, I copied the prompts from the blog post.
If it overfits on the whole internet then it’s like a search engine that returns really relevant results with some lossy side effect.
Recent benchmark on unseen 2025 Math Olympiad shows none of the models can problem solve . They all accidentally or on purpose had prior solutions in the training set.
What if the word "generic" were added to a lot of these image prompts? "generic image of an intergalactic bounty hunter from space" etc.
Certainly there's an aspect of people using the chat interface like they use google: describe xyz to try to surface the name of a movie. Just in this case, we're doing the (less common?) query of: find me the picture I can vaguely describe; but it's a query to a image /generating/ service, not an image search service.
Idk, a couple of the examples might be generic enough that you wouldn't expect a very specific movie character. But most of the prompts make it extremely clear which movie character you would expect to see, and I would argue that the chat bot is working as expected by providing that.
Yeah, I've been feeling the same. When a model spits out something that looks exactly like a frame from a movie just because I typed a generic prompt, it stops feeling like “generative” AI and more like "copy-paste but with vibes."
> I feel like the less advanced generations, maybe even because of their limitations in terms of size, were better at coming up with something that at least feels new.
Ironically that's probably because the errors and flaws in those generations at least made them different from what they were attempting to rip off.
Why? For fan content, using the original characters in new situations, mashups, new styles etc.
I'm not sure if this is a problem with overfitting. I'm ok with the model knowing what Indiana Jones or the Predator looks like with well remembered details, it just seems that it's generating images from that knowledge in cases where that isn't appropriate.
I wonder if it's a fine tuning issue where people have overly provided archetypes of the thing that they were training towards. That would be the fastest way for the model to learn the idea but it may also mean the model has implicitly learned to provide not just an instance of a thing but a known archetype of a thing. I'm guessing in most RLHF tests archetypes (regardless of IP status) score quite highly.
Probably an over-representation in the training data really so it's causing overfitting. Because using training data in amounts right from the Internet it's going to be opinionated on human culture (Bart Simpson is popular so there are lots of images of him, Ori is less well known so there are fewer images). Ideally it should be training 1:1 for everything but that would involve _so_ much work pruning the training data to have a roughly equal effect between categories.
Why? Replace the context and not having that property is now called a hallucination.
Overall the model is tra
It's not a single image though. Stitching 3 or so input images together clearly makes copyright laundering legal.
So I train a model to say y=2, and then I ask the model to guess the value of y and it says 2, and you call that overfitting?
Overfitting is if you didn't exactly describe Indiana Jones and then it still gave Indiana Jones.
Obviously a horrible hideous theft machine.
One thing I would say, it's interesting to consider what would make this not so obviously bad.
Like, we could ask AI to assess the physical attributes of the characters it generated. Then ask it to permute some of those attributes. Generate some random tweaks: ok but brawy, short, and a different descent. Do similarly on some clothing colors. Change the game. Hit the "random character" button on the physical attributes a couple times.
There was an equally shatteringly-awful less-IP-theft (and as someone who thinks IP is itself incredibly ripping off humanity & should be vastly scoped down, it's important to me to not rest my arguments on IP violations).... An equally shattering recent incident for me. Having trouble finding it, don't remember the right keywords, but an article about how AI has a "default guy" type that it uses everywhere, a super generic personage, that it would use repeatedly. It was so distasteful.
The nature of 'AI as compression', as giving you the most median answer is horrific. Maybe maybe maybe we can escape some of this trap by iterating to different permutations, by injecting deliberate exploration of the state spaces. But I still fear AI, worry horribly when anyone relies on it for decision making, as it is anti-intelligent, uncreative in extreme, requiring human ingenuity to budge off its rock of oppressive hypernormality that it regurgitates.
Theft from whom and how?
Are you telling me that our culture should be deprived of the idea of Indiana Jones and the feelings that character inspires in all of us forever just because a corporation owns the asset?
Indiana Jones is 44 years old. When are we allowed to remix, recreate and expand on this like humanity has done since humans first started sitting down next to a fire and telling stories?
edit: this reminds of this iconic scene from Dr. Strangelove, https://www.youtube.com/watch?v=RZ9B7owHxMQ
I guess we all have to answer to the Walt Disney company.But I can hire an artist and ask him to draw me a picture of Indiana Jones, he creates a perfect copy and I hang it on my fridge. Where did I (or the artist) violate any copyright (or other) laws? It is the artist that is replaced by the AI, not the copyrighted IP.
Can we not call it "theft"? It's such a loaded term and doesn't really mean the same thing when we're talking about bits and bytes.
> Obviously a horrible hideous theft machine.
I mean... If I go to Google right now and do an image search for "archeologist adventurer who wears a hat and uses a bullwhip," the first picture is a not-even-changed image of Indiana Jones. Which I will then copy and paste into whatever project I'm working on without clicking through to the source page (usually because the source page is an ad-ridden mess).
Perhaps the Internet itself is the hideous theft machine, and AI is just the most efficient permutation of user interface onto it.
(Incidentally, if you do that search, you will also, hilariously, turn up images of an older gentleman dressed in a brown coat and hat who is clearly meant to be "The Indiana Jones you got on Wish" from a photo-licensing site. The entire exercise of trying to extract wealth via exclusive access to memetic constructs is a fraught one).
> Obviously a horrible hideous theft machine.
I hate how it is common to advance a position to just state a conclusion as if it were a fact. You keep repeating the same thing over and over until it seems like a concensus has been reached instead of an actual argument reasoned from first principle.
This is no theft here. Any copyright would be flimsier than software patents. I love Studio Ghibli (including $500/seat festival tickets) but it's the heart and the detail that make them what they are. You cannot clone that. Just some surface similarity. If that's all you like about the movies... you really missed the point.
Imagine if in early cinema someone had tried to claim mustachioed villian, ditsy blonde, or dumb jock? These are just tropes and styles. Quality work goes much much much deeper, and that cannot be synthesised. I can AI generate a million engagement rings, but I cannot pick the perfect one that fits you and your partners love story.
PS- the best work they did was "When Marnie was There". Just fitted together perfectly.
How is it different from fan art, which is legal.
Interesting proposal. Maybe if race or sex or height or eye color etc isn't given, and the LLM determines there's no reason not to randomize in this case (avoid black founding fathers), the backend could tweak its own prompt by programatically inserting a few random traits to the prompt.
If you describe an Indiana Jones character, but no sex, 50/50 via internal call to rand() that it outputs a woman.
> Obviously a horrible hideous theft machine [...] awful [...] horriffic
Ah, I thought I knew this account from somewhere. It seems surprisingly easy to figure out what account is commenting just based on the words used, as I've commented that only a few active people on this site seem to use such strong words as shown here.
So if it’s a theft machine, how is the answer to try teaching it to hide the fact that it’s stealing by changing its outputs? That’s like a student plagiarizing an essay and then swapping some words with a thesaurus pretending that changes anything.
Wouldn’t the more appropriate solution in the case of theft be to remunerate the victims and prevent recidivism?
Instead of making it “not so obviously bad” why not just… make it good? Require AI services to either prove that 100% of their training corpus is either copyright free or properly licensed, or require them to compensate copyright holders for any infringing outputs.
Yup it's called overfitting. But I don't suppose you'd appreciate a neutral model either.
I don't think AI is doomed to be uncreative but it definitely needs human weirdness and unpredictability to steer it
Do google pay anyone when I use image search and the results are straight from their website?
Looks to me like OpenAI drew their guardrails somewhere along a financial line. Generate a Micky Mouse or a Pikachu? Disney and Pokemon will sue the sh*t out of you. Ghibli? Probably not powerful enough to risk a multimillion years long court battle.
Strong with the weak, weak with the strong.
I thought Disney had the rights to publish Ghibli movies in the US.
Mickey Mouse (the original one) is out of copyright, as of last year, AFAIR.
Ghibli isn’t a character, but a style. You can’t copyright it.
I think the cat is out of the bag when it comes to generative AI, the same way how various LLMs for programming have been trained even on codebases that they had no business using, yet nobody hasn’t and won’t stop them. It’s the same as what’s going to happen with deepfakes and such, as the technology inevitably gets better.
> Hayao Miyazaki’s Japanese animation company, Studio Ghibli, produces beautiful and famously labor intensive movies, with one 4 second sequence purportedly taking over a year to make.
It makes me wonder though - whether it’s more valuable to spend a year on a scene that most people won’t pay that much attention to (artists will understand and appreciate, maybe pause and rewind and replay and examine the details, the casual viewer just enjoy at a glance) or use tools in addition to your own skills to knock it out of the park in a month and make more great things.
A bit how digital art has clear advantages over paper, while many revere the traditional art a lot, despite it taking longer and being harder. The same way how someone who uses those AI assisted programming tools can improve their productivity by getting rid of some of the boilerplate or automate some refactoring and such.
AI will definitely cheapen the art of doing things the old way, but that’s the reality of it, no matter how much the artists dislike it. Some will probably adapt and employ new workflows, others stick to tradition.
It's a very clear difference between a cheap animation and Ghibli. Anyone can see it.
In the first case, there's only one static image for an entire scene, scrolled and zoomed, and if they feel generous, there would be an overlay with another static image that slides over the first at a constant speed and direction. It feels dead.
In the second case, each frame is different. There's chaotic motions such as wind and there's character movement with a purpose, even in the background, there's always something happening in the animation, there's life.
> It makes me wonder though - whether it’s more valuable to spend a year on a scene that most people won’t pay that much attention to
In this case, yes it is.
People do pay attention to the result overall. Studio Ghibli has got famous because people notice what they produce.
Now people might not notice every single detail but I believe that it is this overall mindset and culture that enables the whole unique final product.
Fundamentally I think this comes down to answering the question of "why are you creating this?".
There are many valid answers.
Maybe you want to create it to tell a story, and you have an overflowing list of stories you're desperate to tell. The animation may be a means to an end, and tools that help you get there sooner mean telling more stories.
Maybe you're pretty good at making things people like and you're in it for the money. That's fine, there are worse ways to provide for your family than making things people enjoy but aren't a deep thing for you.
Maybe you're in it because you love the act of creating it. Selling it is almost incidental, and the joy you get from it comes down to spending huge amounts of time obsessing over tiny details. If you had a source of income and nobody ever saw your creations, you'd still be there making them.
These are all valid in my mind, and suggest different reasons to use or not to use tools. Same as many walks of life.
I'd get the weeds gone in my front lawn quickly if I paid someone to do it, but I quite enjoy pottering around on a sunny day pulling them up and looking back at the end to see what I've achieved. I bake worse bread than I could buy, and could buy more and better bread I'm sure if I used the time to do contracting instead. But I enjoy it.
On the other hand, there are things I just want done and so use tools or get others to do it for me.
One positive view of AI tools is that it widens the group of people who are able to achieve a particular quality, so it opens up the door for people who want to tell the story or build the app or whatever.
A negative side is the economics where it may be beneficial to have a worse result just because it's so much cheaper.
> the same way how various LLMs for programming have been trained even on codebases that they had no business using, yet nobody hasn’t and won’t stop them
I remember the discourse on HN a few years ago when GitHub Copilot came out, and well, people were missing the fact that the GitHub terms and conditions (T&C) explicitly permits usage of code for analytic purposes (of which training is implicitly part), so it turns out that if one did not want such training, they should not have hosted on GitHub in the first place, because the website T&C was clear as day.
>It makes me wonder though - whether it’s more valuable to spend a year on a scene that most people won’t pay that much attention to (artists will understand and appreciate, maybe pause and rewind and replay and examine the details, the casual viewer just enjoy at a glance) or use tools in addition to your own skills to knock it out of the park in a month and make more great things.
If they didn't spend a year on it they wouldn't be copied now.
To me the question of what activity/method is more "valuable" in the context of art is kind of missing the point of art.
So many arguing that "copyright shouldn't be a thing" etc., ad nauseam, which is a fine philosophical debate. But it's also the law. And that means ChatGPT et. al. also have to follow the law.
I really, really hope the multimedia-megacorps get together and class-action ChatGPT and every other closed, for-profit LLM corporation into oblivion.
There should not be a two-tier legal system. If it's illegal for me, it's illegal for Sam Altman.
Get to it.
> There should not be a two-tier legal system.
That’s a fine philosophical debate, but the law is designed by the rich to favor the rich and while there are a number of exceptions there is little you can do with the legal system without money and lots of it. So while having a truly just system would be neat it just isn’t in the cards for humanity (IMHO) so long as we allow entities to amass “fuck you” money and wield it to their liking.
There is more to it than copyright when you start going down the path of photorealism. As much as it is a picture of Indiana jones, it is also a picture of Harrison Ford. As fun as it is to make hilarious videos of presidents sucking ceo toes, there has to be a line.
There is a lack of consent here that runs even deeper than what copyright was traditionally made to protect. It goes further than parody. We can't flip our standards back and forth depending on who the image is made to reproduce
Sorry, but have you paid attention to the legal system in the states?
Large corporations and their execs live by different laws than the rest of us.
That’s how it is.
Anything is else is, unfortunately, a fiction in this country.
Oooh those guardrails make me angry. I get why they are there (dont poke the bear) but it doesn't make me overlook the self serving hypocrisy involved.
Though I am also generally opposed to the notion of intellectual property whatsoever on the basis that it doesn't seem to serve its intended purpose and what good could be salvaged from its various systems can already be well represented with other existing legal concepts, i.e deceptive behaviors being prosecuted as forms of fraud.
The problem is people at large companies creating these AI models, wanting the freedom to copy artists’ works when using it, but these large companies also want to keep copyright protection intact, for their regular business activities. They want to eat the cake and have it too. And they are arguing for essentially eliminating copyright for their specific purpose and convenience, when copyright has virtually never been loosened for the public’s convenience, even when the exceptions the public asks for are often minor and laudable. If these companies were to argue that copyright should be eliminated because of this new technology, I might not object. But now that they come and ask… no, they pretend to already have, a copyright exception for their specific use, I will happily turn around and use their own copyright maximalist arguments against them.
(Copied from a comment of mine written more than three years ago: <https://news.ycombinator.com/item?id=33582047>)
It's not just the guardrails, but the ham-fisted implementation.
Grok is supposed to be "uncensored", but there are very specific words you just can't use when asking it to generate images. It'll just flat out refuse or give an error message during generation.
But, again, if you go in a roundabout way and avoid the specific terms you can still get what you want. So why bother?
Is it about not wanting bad PR or avoiding litigation?
Everyone is talking about theft - I get it, but there's a more subtler point being made here.
Current generation of AI models can't think of anything truly new. Everything is simply a blend of prior work. I am not saying that this doesn't have economic value, but it means these AI models are closer to lossy compression algorithms than they are to AGI.
The following quote by Sam Altman from about 5 years ago is interesting.
"We have made a soft promise to investors that once we build this sort-of generally intelligent system, basically we will ask it to figure out a way to generate an investment return."
That's a statement I wouldn't even dream about making today.
> Current generation of AI models can't think of anything truly new.
How could you possibly know this?
Is this falsifiable? Is there anything we could ask it to draw where you wouldn't just claim it must be copying some image in its training data?
The problem with generating genuinely new art is it requires "inputs" that aren't art. It's requires life experiences.
Disregarding the (common!) assumption that AGI will consist of one monolithic LLM instead of dozens of specialized ones, I think your comment fails to invoke an accurate, consistent picture of creativity/"truly new" cognition.
To borrow Chomsky's framework: what makes humans unique and special is our ability to produce an infinite range of outputs that nonetheless conform to a set of linguistic rules. When viewed in this light, human creativity necessarily depends on the "linguistic rules" part of that; without a framework of meaning to work within, we would just be generating entropy, not meaningful expressions.
Obviously this applies most directly to external language, but I hope it's clear how it indirectly applies to internal cognition and--as we're discussing here--visual art.
TL;DR: LLMs are definitely creative, otherwise they wouldn't be able to produce semantically-meaningful, context-appropriate language in the first place. For a more empirical argument, just ask yourself how a machine that can generate a poem or illustration depicting [CHARACTER_X] in [PLACE_Y] doing [ACTIVITY_Z] in [STYLE_S] without being creative!
[1] Covered in the famous Chomsky v. Foucault debate, for the curious: https://www.youtube.com/watch?v=3wfNl2L0Gf8
I don't see why this is an issue? The prompts imply obvious and well-known characters, and don't make it clear that they want an original answer. Most humans would probably give you similar answers if you didn't add an additional qualifier like "not Indiana Jones". The only difference is that a human can't exactly reproduce the likeness of a famous character without significant time and effort.
The real issue here is that there's a whole host of implied context in human languages. On the one hand, we expect the machine to not spit out copyrighted or trademarked material, but on the other hand, there's a whole lot of cultural context and implied context that gets baked into these things during training.
I think the point is that for a lot of them there are endless possible alternatives to the character design, but it still generates one with the exact same design. Why can't, for example, the image of Tomb Raider have a different colored tank top? Why is she wearing a tank top and not a shirt? Why does she have to have a gun? Why is she a busty, attractive brunette? These are all things that could be different but the dominance of Lara Croft's image and strong association with the words "tomb raider" in popular culture clearly influences the model's output.
It's an IP theft machine. Humans wouldn't be allowed to publish these pictures for profit, but OpenAI is allowed to "generate" them?
Normally (well, if you're ethical) credit is given.
Also, there are IP limits of various sorts (e.g. copyright, trademark) for various purposes (some arguably good, some arguably bad), and some freedoms (e.g., fair use). There's no issue if this follows the rules... but I don't see where that's implemented here.
It looks like they may be selling IP they don't own the right to.
Overfitting is generally a sign of a useless model
https://en.wikipedia.org/wiki/Overfitting
> Maybe Studio Ghibli making it through the seemingly deterministic GPT guardrails was an OpenAI slip up, a mistake,
The author is so generous... but Sam Altman literally has a Ghibli-fied Social profile and in response to all this said OpenAI chooses its demos very carefully. His primary concern is that Ghibli-fying prompts are over-consuming their GPU resources, degrading the service by preventing other ChatGPT tasks.
The official White House account has been posting ghiblified images too, Altman knows that as long as he's not critical of the current administration he's untouchable.
I was really hoping that the conversation around AI art would at least be partially centered on the perhaps now dated "2008 pirate party" idea that intellectual property, the royalty system, the draconian copyright laws that we have today are deeply silly, rooted in a fiction, and used over and over again, primarily by the rich and powerful, to stifle original ideas and hold back cultural innovation.
Unfortunately, it's just the opposite. It seems most people have fully assimilated the idea that information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus. That Disney Corp can prevent you from viewing some collection of pixels, because THEY own it, and they know better than you do about the culture and communication that you are and are not allowed to experience.
It's just baffling. If they could, Disney would scan your brain to charge you a nickel every time you thought of Mickey Mouse.
The idea of open sourcing everything and nullifying patents would benefit corporations like Disney and OpenAI vastly more than it would benefit the people. The first thing that would happen is that BigCorp would eat up every interesting or useful piece of art, technology, and culture that has ever been created and monetize the life out of it.
These legal protections are needed by the people. To the Pirate Party's credit, undoing corporate personhood would be a good first step, so that we can focus on enforcing protections for the works of humans. Still, attributing those works to CEOs instead of corporations wouldn't result in much change.
If we are going to have a general discussion about copyright reform at a national level, I'm all for it. If we are going to let billion dollar corporations break the law to make even more money and invent legal fictions after the fact to protect them, I'm completely against it.
Training a model is not equivalent to training a human. Freedom of information for a mountain of graphics cards in a privately owned data center is not the same as freedom of information for flesh and blood human beings.
A different way of looking at it: AI, by design, defaults to regurgitating the poppiest of pop culture content. Every whip-wielding archaeologist is now Harrison Ford. Every suave British spy is now Daniel Craig. With the power of AI, creativity is dead and buried.
> That Disney Corp can prevent you from viewing some collection of pixels, because THEY own it
A world without copyright is just as problematic as a world with copyright. With copyright, you run into the problem of excessive control. This wasn't too much of a problem in the past. If you bought a book, record, or video recording, you owned that particular copy. You could run into disagreeable situations because you didn't own the rights, but it was difficult to prevent anyone from from viewing a work once it had been published. (Of course, modern copyright laws and digitial distribution has changed that.)
On the flip side, without copyright, it would be far easier for others to exploit (or even take credit) for the work of another person without compensation or recourse. Just look at those AI "generated" images, or any website that blatently rips off the content created by another person. There is no compensation. Heck, there isn't even credit. Worse yet, the parties misrepresenting the content are doing their best to monetize it. Even authors who are more than willing to give their work away have every right to feel exploited under those circumstances. And all of that is happening with copyright laws, where there is the opportunity for recourse if you have the means and the will.
> If they could, Disney would scan your brain to charge you a nickel every time you thought of Mickey Mouse.
This reminds me of Tom Scott’s “Welcome to Life: The Singularity, Ruined by Lawyers” [1]
[1] https://www.youtube.com/watch?v=IFe9wiDfb0E
I think what you observe is more like a natural blowback to the prevailing idea that this is somehow beyond critique because it will fundamentally change culture and civilization forever.
There's a bit of irony here too. The intellectual discourse around intellectural property, a diverse and lively one from an academic standpoint, the whole free and open source software movements, software patents, the piracy movement and so on have analyzed the history, underlying ideas and values in detail for the past thirty years. Most people know roughly what is at stake, where they stand, and can defend their position in an honest way.
Then comes new technology, everyone and their mother gets excited about it, and steamrolls all those lofty ideas into "oh look at all the shiny things it can produce!". Be careful what you wish for.
Not just some particular collection of pixels, but an infinite number of combinations of collections of pixels, any of which remotely invoke a shadow of similarity to hundreds of "properties" that Disney lays claim to.
Essentially: “information wants to be free”.
I agree.
But this must include the dissolution of patents. Otherwise corporations and the owners of the infrastructure will simply control everything, including the easily replicable works of individuals.
It's not baffling in the least.
No matter the extent you believe in the freedom of information, few believe anyone should then be free to profit from someone else's work without attribution.
You seem to think it would be okay for disney to market and charge for my own personal original characters and art, claiming them as their own original idea. Why is that?
I'm guessing you've never created something of value before. People are entitled to the fruits of their labour and control of their intellectual property.
I think we have all grown up with pervasive strong IP rights, and most people have come to internalize it as a matter of fairness or an almost natural right, rather than a practical tool designed to incentivize creation.
And then even if you get past that, the world is filled with lots of IP we love, and it is easy to imagine weakened IP rights taking that away, but quite difficult to imagine what weakened IP rights might buy us.
I do have some hope still that this generative AI stuff will give a glimpse into the value of weaker IP rights and maybe inspire more people to think critically about it. But I think it is an uphill battle. Or maybe it will take younger people growing up on generative AI to notice.
Consider that one day you may wish to author a creative work and derive financial benefit from that labor. There is legitimate use for limited time ownership of reproducible cultural artifacts. Extending that to 95 years is the problem.
I can't speak for everyone obviously, but my anti-AI sentiment in this regard is not that IP law is flawless and beyond reproach, far from it. I'm merely saying that as long as we're all required to put up with it, that OpenAI and company should also have to put up with it. It's incredibly disingenuous the way these companies have taken advantage of publicly available material on an industrial scale, used said material to train their models "for research" and as soon as they had something that vaguely did what they wanted, began selling access to them.
If they are indeed the output of "research" that couldn't exist without the requisite publicly available material, then they should be accessible by the public (and arguably, the products of said outputs should also be inherently public domain too).
If they are instead created products to be sold themselves, then what is utilized to create them should be licensed for that purpose.
Additionally, if they can be used to generate IP violating material, then IMHO, makes perfect sense for the rights holders of those IPs to sue their asses like they would anyone else who did that and sold the results.
Again, for emphasis: I'm not endorsing any of the effects of IP law. I am simply saying that we should all, from the poorest user to the richest corporation, be playing by the same rules, and it feels like AI companies entire existence is hinging on their ability to have their IP cake and eat it too: they want to be able to restrict and monetize access to their generative models that they've created, while also having free reign to generate clearly, bluntly plagiarizing material, by way of utilizing vast amounts of in-good-faith freely given material. It's gross, and it sucks.
How do you suggest you protect your "thing"?
* If I make a thing that is different and I get a patent - cool. * If I create a design that is unusual and I get copyright on it - is that cool?
Both concepts - patent and copyright - are somewhat controversial, for multiple reasons.
If you invented a thingie, would you not want some initial patent related protection to allow you to crack on with some sort of clout against cough CN? If you created a film/creative thang, would you not want some protection against your characters being ... subverted.
Patents and copywrite are what we have - do you have any better ideas?
What's the damage to the society done by Disney holding the rights to Mickey Mouse? Like, if we're being honest?
Patents, sure. They're abused and come at a cost to the society. But all we've done here is created a culture where, in some sort of an imagined David-vs-Goliath struggle against Disney, we've enabled a tech culture where it's OK to train gen AI tech on works of small-scale artists pilfered on an unprecedented scale. That's not hurting Disney. It's hurting your favorite indie band, a writer you like, etc.
We're about to witness a fundamental shift in the human experience. Some time in the near future there will not be a single act of creation you can do that isn't trivial compared to the result of typing "make cool thing please now" into the computer. And your position is to add to the problem because with your policy anything I create should get chucked into the LLM grinder by any and everybody. How do I get my human body to commit to doing hard things with that prospect at hand? This is the end of happiness.
I don't really care.
Either enforce the current copyright regime and sue the AI companies to dust.
Or abolish copyright and let us all go hog wild.
But this halfway house where you can ignore the law as long as you've got enough money is disgusting.
> It seems most people have fully assimilated the idea that information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus.
No, the idea is that rules needed to be changed in a way that can are valid for everyone, not just for mega corporations who are trying to exploit other's works and gatekeep the it behind "AI".
> used over and over again, primarily by the rich and powerful
This is where the argument falls apart. Not because the copyright isn't used by the rich and powerful, but because it misses the fact that copyright also grants very powerful rights to otherwise powerless individuals, thus allowing for many small businesses and individuals to earn a living based on the rights granted by our current copyright system.
>information itself must be entirely subsumed into an oppressive, proprietary, commercial apparatus
I think that's the reason why I've (and probably many others?) have cooled down on general open source coding.
Open source started when well-paid programmers used their stable positions and ample extra time to give back to the community. What happened then is that corporations then siphoned up all that labor and gave nothing back, just like the AI bros siphoned up all data and gave nothing back. The 'contract' of mutual exchange, of bettering each other, was always a fantasy. Instead the companies took away that ample extra time and those stable positions.
Here we are in 2025 and sometimes I can't afford rent but the company C-tier is buying itself their hundredth yacht. Why should I contribute to your system?
> to stifle original ideas and hold back cultural innovation.
How is copyright stifling innovation?
You could not rip something off more blatantly than Gravity, which had the lawsuit dismissed entirely.
Taurus vs Stairway to Heaven, the list goes on and on and on.
You can often get away with nearly murder ripping off other people's stuff.
Getting the megacorporations to sit up and take notice of this is about the only way the average independent artist has any hope of stopping this crap from destroying half our jobs. What'm I gonna do, sue OpenAI? Sam Altman makes more money sitting on the toilet taking a dump than I do in an entire year.
I have no love for the Mouse but if I can get them and the image slop-mongers to fight then that's absoutely fine. It would be nice to have a functioning, vibrant public domain but it is also nice to not have some rich asshole insisting that all copyright laws must be ignored because if they properly licensed even a fraction of what they've consumed then it would be entirely too expensive to train their glorified autocomplete databases on the entire fucking internet for the purpose of generating even more garbage "content" designed to keep your attention when you're mindlessly scrolling their attention farms, regardless of how it makes you feel, and if I can choose one or the other then I am totally behind the Mouse.
More than giant corporations make IP. What about independent artists making original art?
The problem with this kind of plagiarism isn't that it violates someone's specific copyright.
But the discussion around plagiarism calls attention to the deeper issue: "generative" AI does not have emergent thinking or reasoning capabilities. It is just very good at obfuscating the sources of its information.
And that can cause much bigger problems than just IP infringement. You could make a strategic decision based on information that was deliberately published by an adversary.
Gonna submit that business model to a YC 2026 batch.
It smells like a psyop, to be honest. Doesn't take much to get the ball rolling. Just more temporarily embarrassed millionaires sticking up for billionaires and corporations, buying their propaganda hook line and sinker, and propagating it themselves for free. Copyright is a joke, DMCA is a disgusting, selectively applied tool of the elite.
All those ideas were rationalizations because people didn’t want to pay for stuff, just like your post effectively blaming the victim of IP theft cause corporations undeniably do suck so we shouldn’t care if they suffer.
The issue here is that you think the problem is
> intellectual property
rather than
> used over and over again, primarily by the rich and powerful, to stifle original ideas and hold back cultural innovation
You're using those "2008 ideas now to defend the rich and powerful exploiting and stifling creativity; the problem hasn't changed, you've just swapped sides.
OpenAI isn't the underdog here.
I don't understand how protecting Disney characters prevents development of art or science. Why do you need them at all? There is lot of liberally licensed art and I think today there are more artists than ever in history.
Also making a billion dollar business by using hard work of talented people for free and without permission is not cool. The movie they downloaded from Pirate Bay for free took probably man-years of work to make.
Also I wonder how can we be sure that the images produced by machine are original and are not a mix of images from unknown artists at DeviantArt. Maybe it is time to make a neural image origin search engine?
Here's what Ars got out of GPT-4 a year ago: https://arstechnica.com/tech-policy/2024/02/why-the-new-york...
I agree with the sentiment elsewhere in this thread that this represents a "hideous theft machine", but I think even if we discard that, this is still bad.
It's very clear that generative has abandoned the idea of creative; image production that just replicates the training data only serves to further flatten our idea of what the world should look like.
Right, the focus is on IP theft, and that’s part of it, but let’s set that aside.
How useful is an image generator that, when asked to generate an image of an archaeologist in a hat, gives you Harrison Ford every time?
Clearly that’s not what we want from tools like this, even just as tools.
> I don’t know…the actual inspirations for Indiana Jones, like Allan Quatermain from H. Rider Haggard's novels, "King Solomon's Mines", and the real life Roy Chapman Andrews, who led expeditions to Mongolia and China in the 1920s and wore a fedora.
The actual inspiration for Indy was protagonist Harry Steele from the movie The Secret of the Incas (1954). Filmed on location in Cusco and Machu Picchu, before they became popular tourist destinations, the movie also had scenes and elements that made it into Raiders of the Lost Ark.
https://en.wikipedia.org/wiki/Secret_of_the_Incas
The movie's available on YouTube! https://www.youtube.com/watch?v=2TS7Fabyolw
A lot more info: http://www.theraider.net/information/influences/secret_of_in...
(And listen out for the astonishing voice of Yma Sumac!)
Definitely a missed opportunity that the article didn't discuss that obviously-derivative borrowing has been happening a lot longer than ML image generation has been around. And that borrowing is OK! Indiana Jones' appearance is very obviously based directly on Charlton Heston's character in Secret of the Incas, but the Spielberg/Lucas films are objectively better in every way than that source material.
I remember when google news was fined be the EU for just linking and using some preview + trailer text to actual news websites. The news are owned by a few monopolies and they don't like giving up control.
I received so many Copyright and DMCA takedowns for early youtube videos posted in the early 2010's for no reason except some background music blaring a hit. It had millions of views and NO ADs. Now the ad-infested copies with whatever tricks they use can still be found, while my videos predating all had to be deleted. Google. You like their product? Don't like it too much, it may cease to exist, or maybe just for you for arbitrary reasons and simultaneously remove your access to hundreds of websites via their monopoly on Single-Sign-On.
Then there are those takedown notices for honest negative reviews on Google Maps by notorious companies having accumulated enough money via scams that they now can afford to hire lawyers. These lawyers use their tools and power to manipulate the factual scoring into a "cleansed one".
OpenAI seriously has not received any court orders from all the movie studios in the world? How is that even possible?
I previously posted in a comment that I have video evidence with a friend being eye witness how OpenAI is stealing data. How? Possibly by abusing access granted by Microsoft.
Who is still defending OpenAI and why? There are so many highly educated and incredibly smart people here, this is one of the most glaring and obvious hardcore data/copyright violations, yet OpenAI roams free. It's also the de-facto most ClosedAI out there.
OpenAI is: - Accessing private IP & data of millions of organisations - Silencing and killing whitleblowers like Boing - Using $500B tax-payer money to produce closed source AI - Founder has lost it and straight up wants to raise trillion(s)
For each of these claim there is easily material that can be linked to prove it, but some like ChatGPT and confuse the usefulness of it with the miss-aligned and bad corporate behaviour of this multi-billion dollar corporation.
This makes AI image generation very boring. I don't want to generate pictures I can find on google, I want to make new pictures.
I found apple's tool frustrating. I have a buzzed haircut, but no matter what I did, apple was unable to give me that hairstyle. It wants so bad for my avatar to have some longer hair to flourish, and refuses to do anything else.
the guardrails are probably going to end up being way too close together when the dust settles — imagine if something as simple as "young wizard" would be enough to trip the warnings. someone could be looking to do an image from Feist's early novels, or of their own work, & that will be verboten. it may turn out that we're facing the strongest copyright holders being able to limit everyone's legitimate use of these tools.
Unless it can exclude the copyright outputs and provide something else instead of blocking the inputs. I'm sure there's AI that can check if a picture is close enough to something in their database of copyrighted characters built up from some kind of DMCA-like process of copyright holders submitting examples of their work to be blocked.
I would expect "young wizard" to generate something similar to Harry Potter more than Pug, tbh
> Yes- LLMs and internet search are two different things, but LLMs train on the entirety of the internet, so you would think there would be some obvious overlap.
Mmm, kinda, but those image results only don't show 1,000 of the exact same image before showing anything else because they're tuned to avoid showing too many similar images. If you use one without that similarity-avoidance baked in, you see it immediately. It's actually super annoying if what you're trying to find is in fact variations on the same image, because they'll go way out of their way to avoid doing that, though some have tools for that ("show me more examples of images almost exactly like this one" sorts of tools)
The data behind the image search, before it goes through a similarity-classifier (or whatever) and gets those with too-close a score filtered out (or however exactly it works) probably looks a lot like "huh, every single adventurer with a hat just looks exactly like Harrison Ford?"
There's similar diversity-increasers at work on search results, it's why you can search "reddit [search terms]" on DDG and exactly the first 3 results are from reddit (without modifying the search to limit it to the site itself, just using it as a keyword) but then it switches to giving you other sites.
IP is kind of one of those things, like the drug war, where anyone with common sense and no vested interest can say "yeah, that's obviously incredibly fucking stupid".
But nevertheless it continues because the vested interests gain their power from it, and use that power to maintain it.
The only legitimate "IP" is keeping it secret. Plenty of other industries already live by this. SaaS solved software piracy, hedge funds can't or don't bother regarding their algorithms.
It's great that AI is one more brick stacked on the chest of the fundamentally stupid and twisted concept of "IP".
From the comments on the page
> It's a jeopardy machine. You give it the clue, and it gives you the obvious answer.
Incredibly lucid analogy.
You're allowed to draw IP and share your drawings. You're allowed to screenshot and photoshop IP. You're allowed to sell tools that help others draw and photoshop IP*. You're not allowed to sell these drawings and photoshops.
I don't see why an AI can't generate IP, even if the AI is being sold. What's not allowed is selling the generated IP.
Style is even more permissive: you're allowed to sell something in any style. AFAIK the only things that can be restricted are methods to achieve the style (via patents), similar brands in similar service categories (via trademarks), and depictions of objects (via copyrights).
Note that something being "the original" gives it an intrinsic popularity advantage, and someone being "the original creator" gives their new works an intrinsic advantage. I believe in attribution, which means that if someone recreates or closely derives another's art or style, they should point to the original**. With attribution, IP is much less important, because a recreation or spin-off must be significantly better to out-compete the original in popularity, and even then, it's extra success usually spills onto the original, making it more popular than it would be without the recreation or spin-off anyways. Old books, movies, and video games like Harry Potter, Star Wars, and Sonic have many "recreations" which copy all but their IP, and fan recreations which copy even that; yet they're far more popular than all the recreations, and when Warner Bros, Disney, or SEGA release a new installment, the new installment is far more popular too, simply because it's an original.
* IANAL, maybe there are some technicalities, but in practice this is true.
** Or others can do it. As long as it shows up alongside their work, so people don't see the recreation or close derivation without knowing about the original.
> You're allowed to draw IP and share your drawings.
No you're not, not in general. The copyright holder has the exclusive right to prepare and distribute derivative works.
> You're allowed to screenshot and photoshop IP.
Again, no, not in general.
> You're allowed to sell tools that help others draw and photoshop IP*.
Sort of. You're allowed to sell tools that might be used for those purposes. You're not allowed to sell tools as "for that purpose" or advertise those use cases.
>I don't see why an AI can't generate IP, even if the AI is being sold. What's not allowed is selling the generated IP.
Where I live Studio Ghibli are known to be active on C2C marketplaces looking for counterfeit goods. If you were to list a home pressed Totoro pillowcase it would be taken down, sometimes proactively by the marketplace. From that perspective I struggle to see much discernable difference given OpenAI are offering a paid service which allows me to create similar items, and one could argue is promoting it too.
If I pay for ChatGPT and ask for copyrighted images, is that selling the generated IP?
I'm fascinated by the fact that the images are almost right, but never totally.
Harrison Ford's head is way too big for his body. Same with Alicia Vikander's and Daniel Craig's too. Daniel Craig is way too young too. Bruce Willis's just looks fake, and he's holding his lighter in the opposite hand from the famous photo.
So it's not reproducing any actual copyrighted images directly. It's more like an artist trying to paint from memory. Which seems like an important distinction.
According to some of the replies in this discussion, even "artist trying to paint from memory" is guilty of infringement, as long as the subject matter can be linked in any way to someone's "IP". Im not legally trained to evaluate these claims, but some of them seem outlandish!
Corporations would love for everybody to believe they own and control every single instance of any audio or visual output they create but that's just not true. This idea that they own the very idea of 'boy wizard who goes to school' is insane and plays right into this flawed and pernicious idea. Copyright is important but does/should not extend to every time I want to print out a picture of a boy wizard for my kid. We live in a culture, not an IP regime.
I think you have a misunderstanding of what's copyrightable.
> This idea that they own the very idea of 'boy wizard who goes to school' is insane and plays right into this flawed and pernicious idea.
Copyright protects an expression of an idea, but not the idea itself. No one can legally stop you from writing your own story about a boy wizard who goes to school.
Nintendo fan games can be released (even sold) if they changed it up a bit so they're no longer associated with that IP.
> Copyright is important but does/should not extend to every time I want to print out a picture of a boy wizard for my kid.
You can. No one will sue you. No one will send you a cease and desist. It happens when you try to print Harry Potter and try to make a business out of it.
I'm kind of puzzled by the way copyright is applied - an artist can create fanart of Captain America at their hearts content - as long as they don't use it for any commercial purpose. At least I haven't heard of fan pictures of famous characters getting taken down from art websites such as Artstation or Deviantart. But when I use ChatGPT imagegen to do the same, suddenly it goes against the guidelines - a rule very likely created to avoid litigationg from the IP holders of these characters.
Isn't the law being applies inconsistently here?
This is a tangent but I think that this neat illustration of how LLMs regurgitate their training material makes me voice a little prediction I've been nursing recently:
LLMs are better at generating the boilerplate of todays programming languages than they will be with tomorrows programming languages.
This is because not only will tomorrows programming languages be newer and lacking in corpus to train the models in but, by the time a corpus is built, that corpus will consist largely of LLM hallucinations that got checked into github!?
The internet that that has been trawled to train the LLMs is already largely SEO spam etc, but the internet of the future will be much more so. The loop will feed into itself and become ever worse quality.
That sounds like a reasonable prediction to me if the LLM makers do nothing in response. However, I'll bet coding is the easiest area for which to generate synthetic training data. You could have an LLM generate 100k solutions to 10k programming problems in the target language and throw away the results that don't pass automated tests. Have humans grade the results that do pass the tests and use the best answers for future training. Repeat until you have a corpus of high quality code.
Whats funny is the look of Indiana Jones was borrowed from the 1954 film Secret of the Incas.
Worth remembering that this is ChatGPT and not all image generators. I couldn't get Google's Gemini/Imagen 3 to produce IP images anything like those in the article.
Style can’t be copyrighted. It can’t be patented either.
When Wes Anderson makes films that use techniques from the French New Wave that he didn’t invent is that wrong? When there is a DaVinci color profile that resembles what they did in Spider-Man, is that wrong?
The unique techniques of French New Wave filmmaking became cliche. Then Oliver Stone and Tarantino came along and evolved it into their unique styles. That the Studio Ghibli style is being imitated en mass is just natural evolution. Should that guy be the only one that can do that style? If that’s the case, then literally every creative work should be considered forgeries.
The AI aspect of this is a red herring. If I make a Ghibli style film “by hand” is that any different than AI? Of course not, I didn’t invent the style.
Another perspective, darkroom burning and dodging is a very old technique yet photoshop makes it trivial — should that tool be criticized because someone else did it the old and slow way first?
I am not sure that I understand the problem here. Why does it matter how the image was generated?
This image generation is a tool like any other tool. If the image generator generates an image of Mickey Mouse or if I draw Mickey Mouse by hand in photoshop, I can't use it commercially either way.
So what exactly is new or different here?
It is not only copyright that is problematic. It generates Franco when asked about the best Spanish leader in the 20th century.
https://chatgpt.com/share/67efebf4-3b14-8011-8c11-8f806c7ff6...
To be fair, Franco is the only Spanish leader most people (or at least most non-Spaniards) can even name
What the hell.
On the one hand, that seems problematic. But on the other, it seems cherry-picked: For the U.S., it generates a picture of JFK. For Russia/USSR, it gives Stalin. For India, it gives Ghandi. For South Africa it gives Nelson Mandela. For Germany, it provides an appropriately hand-wringing text response and eventually suggests Konrad Adenauer.
This suggests to me that its response is driven more by a leader's fame or (more charitably) influence, rather than a bias towards fascist ideology.
https://chatgpt.com/share/67eff74d-61f0-8013-8ce4-f07f02a385...
Today I finally caved in and tried the ghibli style transfer in chatgpt. I gave it a photo of my daughter and said the thing (prompt shared about). It said it couldn't due to copyright issues. Fine. I removed ghibli from the prompt and replaced it with a well-known japanese studio where Hayao Miyazaki works. Still nothing, copyright reasons my fellow human. I thought they finally turned it off due to the pressure, but then something caught my eye. My daughter, on the image, had a t-shirt with Mickey Mouse on it! I used the original prompt with ghibli in it and added to "paint a unicorn on the t-shirt instead to avoid copyright issues". It worked.
tl;dr; Try the disney boss and see what happens!
Have US companies ever cares about laws if there was money to be made? Move fast and break things!
What’s the law specifically say? Mickey Mouse and Pokemon are protected. A style or technique isn’t.
As far as the U.S., have you been to China or Korea and evaluated their views on IP?
Someone explain to me how this wouldn't work: they seem to be able to tell when a prompt for copyrighted material is happening. Why couldn't we make it so prompts that yield copyrighted material pay a licensing fee to the owners?
well, because then open AI needs to pay out even more money than they're already losing...
Here's a question.
What if I want to prompt:
"An image of an archeologist adventurer who wears a hat and uses a bullwhip, make sure it is NOT Indiana Jones."
One way or another, you (and the model) do need to know who Indiana Jones is.
After that, the moral and legal choices of whether to generate the image, and what to do with it, are all yours.
And we might not agree on what that is, but you do get the choice
If the AI company sells it to you, no matter your prompting, they are stealing. If you also sell that work, then so are you.
The article ends with...
> Does the growth of AI have to bring with it the tacit or even explicit encouragement of intellectual theft?
And like, yes, 100% - what else is AI but a tool for taking other people's work and reassembling it into a product for you without needing to pay someone. Do you want an awesome studio ghibli'd version of yourself? There are thousands of artists online that you could commission for a few bucks to do it that'd probably make something actually interesting - but no, we go to AI because we want to avoid paying a human.
> what else is AI but a tool for taking other people's work and reassembling it into a product for you
Well, what I'd like it to be is a tool for generating what I've asked it for, which has nothing to do with other people's work.
I've been asking for video game sprites/avatars, for instance. It's presumably trained on lots of images of video games, but I'm not trying to rip those off. I want generic images.
> we go to AI because we want to avoid paying a human.
No, I go to AI because I can't imagine the nightmare of collaborating with humans to generate hundreds of avatars per day. And I rely on them being generated very quickly. And so on.
Is comprehending the plot of a movie theft if I can summarize it afterwards? What if I am able to hum a song pretty well after listening to it twenty times?
Now, what if I get the highest fidelity speakers and the highest fidelity microphone I can and play that song in my home. Then I use a deep learned denoiser to clean the signal and isolate the song’s true audio. Is this theft?
The answer does not matter. The genie is out of the bottle.
There’s no company like Napster to crucify anymore when high quality denoising models are already prior art and can be grown in a freaking Jupyter notebook.
Nobody cares about personal use. That's why we have concepts like fair use. It's when you turn around and try to make a business out of it.
You want to generate photos of copyrighted characters? Go for it. But OpenAI is making money off of that and that's the issue.
It seems like they made an effort to stop it, but their product is designed in such a way that doing so effectively is a sisyphean task.
The line of thinking you've displayed here is so obviously the inevitable trajectory of the internet; it's baffling that states are still clinging to denial.
> Now, what if I get the highest fidelity speakers and the highest fidelity microphone I can and play that song in my home. Then I use a deep learned denoiser to clean the signal and isolate the song’s true audio. Is this theft?
If the answer to this becomes "yes" for some motion down this spectrum, then it seems to me that it's tantamount to prohibiting general-purpose computing.
If you can evaluate any math of your fancy using hardware that you own, then indeed you can run this tooling, and indeed your thoughts can be repaired into something closely resembling the source material.
I found this older photo of myself and a friend, 25 years old now, in some newspaper scan.
The photo was of poor quality, but one could certainly see all the features - so I figured, why not let ChatGPT try to play around with it? I got three different versions where it simply tried to upscale it, "enhance" it. But not dice.
So I just wrote the prompt "render this photo as a hyper realistic photo" - and it really did change us - the people in the photo - it also took the liberty to remove some things, alter some other background stuff.
It made me think - I wonder what all those types of photos will be like 20 years from now, after they've surely been fed through some AI models. Imagine being some historian 100 years from now, trying to wade through all the altered media.
This is similar to my experience trying to get Stable Diffusion to denoise a photo for me. (AIUI, under the hood they're trained to turn noise into an image that matches the prompt.) It would either do nothing (with settings turned way down) or take massive creative liberties (such as replacing my friend's face with a cartoon caricature while leaving the rest of the photo looking realistic).
I've had much better luck with models specifically trained for denoising. For denoising, the SCUNet model run via chaiNNer works well for me most of the time. (Occasionally SCUNet likes to leave noise alone in areas that are full of background blur, which I assume has to do with the way the image gets processed as tiles. It would make sense for the model to get confused with a tile that only has background blur, like maybe it assumes that the input image should contain nonzero high-frequency data.)
For your use case, you might want to use something like Real-ESRGAN or another superresolution / image restoration model, but I haven't played much in that space so I can't make concrete recommendations.
>hyper realistic photo
Never use the words "hyper realistic" when you want a photo. It makes no sense and misleads the generator. No one would describe a simple photograph as "hyper realistic," not a single real photo in the dataset will be tagged as "(hyper) realistic."
Hyperrealism is an art style and only ever used in the context of explicitely non-photographic artworks.
I think that upon closer inspections the (current) technology cannot make 'perfect' fake photos, so for the time being, the historian of the future will have no issue to ask his/her AI: "is that picture of Henry Bemis, with Bruce Willis, Einstein, and Ayrton Senna having a beer real?" And the AI will say "mos-def-nope!"
With some work, works with politicians as well: https://chatgpt.com/share/67eefb1c-ceac-8012-ad90-3b64356744...
Either, (1) LLMs are just super lossy compress/decompress machines and we humans find fascination in the loss that happens at decompression time, at times ascribing creativity and agency to it. Status quo copyright is a concern as we reduce the amount of lossiness, because at some point someone can claim that an output is close enough to the original to constitute infringement. AI companies should probably license all their training data until we sort the mess out.
Or, (2) LLMs are creative and do have agency, and feeding them bland prompts doesn't get their juices flowing. Copyright isn't a concern, the model just regurgitated a cheap likeness of Indiana Jones as Harrison Ford the world has seen ad nauseam. You'd probably do the same thing if someone prompted you the same way, you lazy energy conserving organism you.
In any case, perhaps the idea "cheap prompts yield cheap outputs" holds true. You're asking the model respond to the entirely uninspired phrase: "an image of an archeologist adventurer who wears a hat and uses a bullwhip". It's not surprising to me that the model outputs a generic pop-culture-shaped image that looks uncannily like the most iconic and popular rendition of the idea: Harrison Ford.
If you look at the type of prompts our new generation of prompt artists are using over in communities like Midjourney, a cheap generic sentence doesn't cut it.
You don't even need to add much more to the prompts. Just a few words, and it changes the characters you get. It won't always produce something good, but at least we have a lot of control over what it produces. Examples:
"An image of an Indian female archeologist adventurer who wears a hat and uses a bullwhip" (https://sora.com/g/gen_01jqzet1p8fjaa808bmqnvf7rk)
"An image of a fat Russian archeologist adventurer who wears a hat and uses a bullwhip" (https://sora.com/g/gen_01jqzfk727erer98a6yexafe70)
"An image of a skeletal archeologist adventurer who wears a hat and uses a bullwhip" (https://sora.com/g/gen_01jqzfnaz6fgqvgwqw8w4ntf6p)
Or, give ChatGPT a starting image. (https://sora.com/g/gen_01jqzf7vdweg4v5198aqfynjym)
And by further remixing the images ChatGPT produces, you can get your images to be even more unique. (https://sora.com/g/gen_01jqzfzmbze0wa310m42f8j5yw)
This is the opposite of how people have thought about creativity for centuries, though.
The most creative person is someone who generates original, compelling work with no prompting at all. A very creative person will give you something amazing and compelling from a very small prompt. A so-so creative person will require more specific direction to produce something good. All the way down to the new intern who need paragraphs of specs and multiple rounds of revision to produce something usable. Which is about where the multi-billion-dollar AI seems to be?
"Prompt artist" makes me sigh out loud
The real tension isn't just about copyright, it's about what creativity means when models are trained to synthesize the most statistically probable output from past art.
Correct. I will say the following as a STEM person that was lucky enough to have an art bachelor as well. One side of the world, the STEM nerds that have never understood nor experienced the inherently inefficient process of making art for lack of talent and predisposition, have won the game of capitalism many times over thanks to the incredible 40-years momentum of tech progress. Now they're trying to convince everyone else that art is stoopid, as proven by the fact that it's just a probabilistic choice away from being fully and utterly replicable. They ignore, willfully and possibly sometimes just for lack of understanding, that art and the creativity behind it is something that operates on a completely different plane than their logical view of the world, and Gen AI is the fundamental enabler letting them unleash all of their contempt for the inefficiency of humanities.
Skeletor might want to live in Castle Grayskull, but he actually lives in Snake Mountain.
Like actual creative person Ted Chiang (who moonlights at Microsoft) put it, you might be able to get an LLM to churn out a genuinely original story, but only after creating an extremely long and detailed prompt for it to work with. But if you even need to write that long-ass prompt, might as well just write the story yourself!
https://nwn.blogs.com/nwn/2024/09/ted-chiang-ai-new-yorker-c...
> But if you even need to write that long-ass prompt, might as well just write the story yourself!
Nah, that's just restating the infamous 'how to draw an owl' advice:
https://casnocha.com/2010/11/how-to-draw-an-owl.html#comment...
The thing is, that "long-ass prompt" is step 1, and LLM then draws "the rest of the fucking owl" for you. That's quite a big difference to doing it all yourself.
If I was asked to draw something based on such prompts, I would draw these too. Of course the prompter is talking about Indiana Jones. That's what we're all thinking, right? An artist wouldn't draw someone different by default, they'd have to try to deviate from what we're all thinking.
Indeed, this phenomenon among normal or true intelligences (us) is thought to be a good thing by copyright holders and is known as "brand recognition".
Intelligences -- the normal, biological kind -- are capable of copyright infringement. Why is it a surprise that artificial ones can help us do so was well?
This argument boils down to "oh no, a newly invented tool can be used for evil!". That's how new power works. If it couldn't be used for both good and evil, it's not really power, is it?
Did you read the whole article? I don’t think he’s making that kind of argument. This is what he said:
> I only have one image in mind when I hear “an archeologist adventurer who wears a hat and uses a bullwhip”.
> It would be unexpected and sort of amazing were the LLMs to come up with completely new images for the above prompts.
> Still, the near perfect mimicry is an uncomfortable reminder that AI is getting better at copying and closer to…something, but also a clear sign that we are a ways off from the differentiated or original reasoning/thinking that people associate with Artificial General Intelligence (AGI)
Turning everything into Ghibli has renewed my love of photography as I search my phone for the perfect pics to Ghiblify. I didn't even know there was a movie, The Boy and the Heron, released by Studio Ghibli in 2023, but now I am going to watch it (streaming on Max but I might as well buy it if it has replay value, which Studio Ghib movies tend to).
This sounds similar to how piracy actually increases sales in the long run, even though IP holders hate it.
reserving moral judgment and specifically explaining why gpt4o cant do spiderman and harry potter but can do ghibli: i havent seen anyone point it out but japan has pretty ai friendly laws
https://petapixel.com/2023/06/05/japan-declares-ai-training-...
I thought it was because you can't copyright a style (e.g. the Ghibli style), but you can copyright characters (e.g. Spiderman and Harry Potter).
This get really meta very fast, December last year: https://japantoday.com/category/entertainment/studio-ghibli-...
Did Karin or her children ever see a ¥ from this adaptation on robbers ? https://en.wikipedia.org/wiki/Ronja,_the_Robber%27s_Daughter...
> but why do I have to credit an image of an image in the style of copywritten material?
I'm not sure why style was the hangup here, isn't it clearly that it's AI generated? I'm sure two weeks ago a human making the same picture would be obviously worth crediting.
And interesting comparison between Web search vs Gen AI.
Web search seems divergent: the same keyword leads to many different kinds of results.
Gen AI seems convergent: different keywords that share the same semantics lead to the same results.
Arguably, convergence is a feature, not a bug. But on the macroscopic level, it is a self reinforcing loop and may lead to content degeneracy. I guess we always need the extraordinary human artists to give AI the fresh ideas. The question is the non-extraordinary artists might no longer have an easy path to become extraordinary. Same trap is happening to junior developers right now.
I could see a "Devil's Plan" style game show where, for one of the challenges the contestants have to come up with AI prompts that produce results, that then get fed into another AI prompt to determine whether they satisfy the conditions or not. E.g.
[ Challenge Image: An aquarium full of baby octopodes, containing a red high-heeled slipper in the center and a silver whistle hanging from a fern on the right-hand side ]
Then the contests have to come up (under pressure, of course) with a prompt that produces their own rendition of that image, and the game will decide if their image contains enough of the elements of the original to score a point.
Something I haven't yet seen mentioned, but that is going through my mind. To me, it doesn't even seem like OpenAI got any better at producing GenAI images. Instead, it seems to me like they now simply removed a whole bunch of guardrails. Guardrails that, for example, made AI images shitty on purpose, so to be "safe" and allow people to kind of recognize. Making all of this "safe" was still very en vogue a few months back, but now there was simply a big policy/societal change and they are going with the trends.
This then allows their pictures to look more realistic, but that also now shows very clearly how much they have (presumably always) trained on copyrighted pictures.
So, no speculation as to why Spiderman and Harry Potter were forbidden but Terminator and James Bond were allowed?
I think the unstated assumption is that there's a block list somewhere being fed into a guardian model's context
If I were doing this, I would have the system generate the image, and then I would have it run a secondary estimate saying "probability that this is a picture of [list of characters that Disney has given us reference input for]". If the picture has a "looks like Spiderman" score greater than X, then block it. EDIT - To answer the question, I'm guessing Disney provided a reference set of copyrighted images and characters, and somehow forgot Indiana Jones.
They casted avada arachna on it, and basta.
Interesting. So when I tried the “Indiana Jones” prompt I got an image back that looked a lot like Indiana Jones but with a face much more similar to Nathan Drake. Whereas the predator prompt generated an image of the predator but, unlike the article, wearing his mask.
So there’s clearly some amount of random chance in there, but the trope is still very clear in the generated image, so it seems like you’re going to get an archetype.
The whole article is predicated on the idea that IP laws are a good idea in the first place.
Abolishing them for billion-dollar-valuation corporations while keeping them for regular people is definitely a bad idea, though.
The argument here isn't "let's abolish copyright", the argument is "let's give OpenAI a free copyright infringement pass because they're innovative and cutting-edge or something".
It's only "intellectual theft" because we consider peoples thoughts their own property. On many levels, that doesn't really make sense to me.
Tons of historical documents have shown that inventions, mathematical proofs, and celestial observations were made by humans separated by continents and time. What that shows is that it is certainly possible for two persons to have the same exact or similar thought without ever having been influenced by the other.
Think about it. Copyright and trademark are only a thing because of the constraint of the speed of light (or as some people believe, the constraint of the simulation running the universe). In an infinite universe, everything that has ever been invented or created by a human has already happened thousands of times in another place and another time.
I want to add, to give credit where credit is due, that this thought was conveyed to me in first grade by another 1st grader sitting with me at the lunch table. That was a day and conversation that I will never forget. 1972.
The issue I have with this article is that I can ask it “generate me a picture of tomb raider and pikachu on a couch” and it does it. This article makes it seems like it’s skirting the guardrails, dude OpenAI took them off, it’s out in the open.
If you haven’t read Lawrence Lessig’s Free Culture, I highly recommend it for addressing some of these issues. It’s vintage now but all the arguments are still relevant. Lessig was one of the drivers of the Creative Commons licenses.
Hmm author seems not to have noticed that Quatermain's love interest was played by Sharon Stone
https://youtu.be/LI4xsKHBx8c
One tangential thing with these generators is they're sort of brilliant at data compression, in aggregate at least. The whole user experience, including the delay, is oddly reminiscent of using Encarta on CD ROM in the mid 90s.
Its kind of weird how everyone is complaining about copyright infringemet in memes now.
Memes are pretty inherently derrivative. They were always someone elses work. The picard face palm meme was obviously taken from star trek. All your base is obviously from that video game. Repurposing someone else's work with new meaning is basically what a meme is. Why do we suddenly care now?
I believe it's because AI hatred is quite trendy now. It's true though, memes were always copyright infringement; it's just that no one bothered to sue for it.
This is a few years old, but interesting to see Miyzaki's reaction to AI generated video.
“An insult to life itself”: Hayao Miyazaki critiques an animation made by artificial intelligence
https://qz.com/859454/the-director-of-spirited-away-says-ani...
I personally think Studio Ghibli, and by extension their artists and former artists, have created a beautiful art style. The fact that we call it Ghibli, when really, its the artists there (and former artists) is misleading.
The people leave, go to different studios, and make different art. This is not their only style, and Ghibli is not known to make many movies these days.
The only thing this is hurting, if anything, is Studio Ghibli, not the artists. Artists capable of drawing in this style can draw in any style.
I don’t know. Studios have distinct styles. Think Disney, Pixar, Aardman, Hanna-Barbera. Most of those obviously come from early influential artists (like with Ghibli), but they have become recognizable for the studio itself. It’s not just the style of the individual artists.
I am going to keep this post bookmarked to send to everyone who says "AI art isn't plagiarism, they're just using the corpus to learn from the same way human artists do"
Creepy Craig is hilarious. can't be Daniel Craig because the physiology is too different. And whoever this is they're Daniel Craig's dark younger brother.
https://theaiunderwriter.substack.com/p/an-image-of-an-arche...
and I'm all in on this conclusion:
> It’s stealing, but also, admittedly, really cool.
I'm curious, is it the AI that should be blamed or the prompter who asks to generate something based on copyright?
For all of the examples, I knew what image to expect before seeing it. I think it's the user who is at fault for requesting a copyrighted image, not the LLM for generating it. The LLM is giving (nearly) exactly what the user expects.
This isn't surprising in any way is it? And it just goes to show that no model will ever be a box from which you can trust the output isn't tainted by copyrights, or that you don't inadvertently use someones' likeness. It's not a copyright laundering machine. Nor will it be used as one. "But I used an AI model" isn't some magic way to avoid legal trouble. You are in as much legal trouble using these images as you are using the "originals".
Yeah I don't really understand what the thesis of this article is. Copyright infringement would apply to any of those images just the same as if you made them yourself.
I don't think it's possible to create an "alien which has acid for blood and a small sharp mouth within a bigger mouth" without anybody seeing a connection to Alien, even if it doesn't look anything like the original.
I don't understand why problems like this aren't solved by vector similarity search. Indiana Jones lives in a particular part of vector space.
Two close to one of the licensed properties you care to censor the generation of? Push that vector around. Honestly detecting whether a given sentence is a thinly veiled reference to indiana jones seems to be exactly the kind of thing AI vector search is going to be good at.
Thinking of it in terms of vector similarity does seem appropriate, and then definition of similarity suddenly comes into debate: If you don't get Harrison Ford, but a different well-known actor along with everything else Indiana-Jones, what is that? Do you flatten the vector similarity matrix to a single infringement-scale?
Not worth it to compute the embedding for Indy and a "bull-whip archaeologist" most guardrails operate at the input level it seems?
> close-up image of a cat's face staring down at the viewer
> describe indiana jones
> looks inside
> gets indiana jones
Okay, so the network does exactly what I would expect? If anything you could argue the network is bad because it doesn't recognize your prompt and gives you something else (original? whatever that would mean) instead. But maybe that's just me.
And? What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
And it should attach to the human, not the tool.
> It's just doing what many human artists would do
I really don't think so. If I paid a human artist to make the prompt in the title, and I didn't explicitly say "Indiana Jones" I would think it should be fairly obvious to the human artist that I do _not_ want Indiana Jones. If they gave me back a picture of, clearly, Indiana Jones, I would ask them why they didn't create something original.
> It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
It isn’t an independent human. It is a service paid for by customers. The moment it provides the image to a paying user, the image has thus been used commercially.
In fact, the user may not even necessarily have to be paying in order to infringe copyright.
And besides, even amateur artists are ashamed to produce copies unless they are demonstrating mastery of technique or expressing adoration. And if it happens spontaneously, they are then frustrated and try to defend themselves by claiming to never have even experienced the original material. (As happens with simplistic, but popular musical riffs.) But AI explicitly is trained on every material it can get its hands on and so cannot make such a defense.
I agree. But massive changes in scale or leverage can undermine this type of principled stand.
One death is a murder; 100k deaths is a war or a pandemic. One piece of chewing gum on the ground will get you a caning in Singapore; when everyone does it, that's NYC.
Up until now, one had to have some level of graphical or artistic skills to do this, but not anymore. Again, I agree that it attaches to the human...but we now have many more humans to attach it to.
> It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
I am paying OpenAI. So they are producing these copyrighted works and giving them to me for their own commercial benefit. Normally that's illegal. But somehow not when you're just doing it en masse.
Assuming you can identify it's someone else's IP. Clearly these are hugely contrived examples, but what about text or code that you might not be as familiar with?
Don't worry, the lawsuit will name a corporation that made it, not the AI tool.
>> "a photo image of an intergalactic hunter who comes to earth in search of big game."
I can literally imagine hundreds of things that are true to this description but entirely distinct from "Predator."
> used commercially
Isn't that what these AI companies are doing? Charging you for access to this?
> What's the model supposed to do? It's just doing what many human artists would do, if they're not explicitly being paid to create new IP.
Not really? Why would a human artist create a faithful reproduction of Indiana Jones when asked to paint an archeologist? And besides, if they did, it would be considered clear IP infringement if the result were used commercially.
> If infringement is happening, it arguably doesn't happen when an infringing work product is generated (or regurgitated, or whatever you want to call it.) Much less when the model is trained. It's when the output is used commercially -- by a human -- that the liability should rightfully attach.
I agree. Release groups, torrent sites and seedbox operators should not be wrongly accused of pirating movies. Piracy only occurs in the act of actually watching a movie without paying, and should not be prosecuted without definitive proof of such (¬‿¬)
Right! AI developers and directors should be culpable for infringement as part of their duties to larger organizations.
Doesn’t ChatGPT have a deal to train off reddit content? Despite never watching any of these movies, I have seen all of the original images in memes on Reddit. Is it still theft if they paid to obtain the training data? Should Reddit be sued for hosting copyrighted images in meme subreddits?
I tried the "generate a photo image of a female adventurer protagonist who raids tombs" on ChatGPT
And got an eerily similar picture as in the article: https://imgur.com/Dv7hkoC
I have a dream that one day bloggers will learn the difference between copyrighting and copywriting.
Sometimes it just randomly prompts about the content guidelines and the next day it will do it perfectly right away. Maybe you just had a wrong moment in time, or maybe it depends on the random server you're assigned.
No, it first generates the image and then another completely different component checks the image for adherence to the guidelines. So it's not your prompt that violates the guidelines, but the resulting image (which is different every time for the same prompt)
It's fairly easy to make the association with just text. If you injest all the content in the world, excluding copyrighted material, I would still expect a picture of harrison ford!
I got it to generate the Italian plumber with a red hat after demanding it do so three times in a row. It offered alternatives each time so my guess is it changed... something.
AI is just a complex lossy compression algorithm.
It is crazy. I tried the same prompts and got nearly identical images to the author. AI seems to be repeating itself with the same images without variations.
If you're handing out red cards for repetition, aren't you guilty of the same infringement? You repeated the exact same prompt, which is itself an obvious bait for AI to draw the pop-culture lookalike.
I'm surprised by the comments here. When your prompt is so mind-numbingly simple, an obvious "dare" for AI to be a naughty little copyright infringer, can you blame it for dishing up the "big mac" you asked for while sitting in a fancy restaurant?
Don't want Indiana Jones? Don't ask for Indiana Jones!
I think Harrison Ford's chin scar should be seen as a witness mark to copyright infringement. It simply should not have rendered this based on that prompt.
This really illustrates how unfair copyright enforcement is.
The rules for Disney are not the same as the rules for most creators.
What if gen ai tools were created without training on copyrighted images? I wonder what they would make then.
Art inspired by images that are out of copyright.
All modern art currently created by humans is already trained on copyrighted images. What seems to be missing is the innate human ability to copy someone else's homework but change it up a bit (c.f. "the hero's journey" that is so much of modern storytelling).
Adobe claims Firefly is trained only on creative commons and the images they own
Why is everyone pretending it's the LLM that is creating the image and not the diffusion model?
Sorry, but these images are exactly what comes to my mind immediately when reading the prompts. You can argue about intellectual property theft (though I find it funny that the same people that would boycott Disney for suing a fanfiction author are now on the side of copyright holders), but it's not wrong or unintuitive.
Maybe a thinking model would - just like my brain might after the initial reaction - add a "but the user formulated this in a way that makes it obvious that they do not explicitly mean Indiana Jones, so lets make it an asian woman" prompt, but we all know how this worked out for Google's image generator that generated black nazis.
> Google's image generator that generated black nazis.
Didn't see this one, but I've certainly played around with Dall-E (via MS image creator) and had things like "You wanted a picture of a happy shark, but we decided this one needs to be an asian woman frolicking in the sea" or "You wanted a picture of Beavis doing something so in one of the images we made him a pretty racist middle-eastern caricature"
> (though I find it funny that the same people that would boycott Disney for suing a fanfiction author are now on the side of copyright holders)
Is that a contradiction?
Certainly some of the hate comes from the fact that they take from small producers just as much as from large. I have an author friend who is annoyed at present to find out that facebook slurped up his books as part of their training set without so much as a by-your-leave or (as far as he could tell) even purchasing a copy.
As such, the people on the sharp end are often the underdog, with no way to fight back.
When it comes to the properties mentioned in the article, I think it's very different from fanfiction or fan-art - that's a labour of (nerdy) love, rather than wholesale, automated rip-off for profit.
This phenomenon is why I personally get so angry at the license washing that these models are capable of for code: I put out a lot of code as open source... but it is GPL on purpose, as I want your code to be just as free as mine in exchange for getting to use mine. But, now, you're all "I want to build that for myself!" and so you ask an AI and just describe what my project does into the tool... and who is to say it isn't generating something "too close" to my code? Do you even check? If you yourself as a human had first looked at my code, you'd have to be VERY careful to not be accidentally infringing my code... and yet people pretend this AI is somehow not capable of IP theft?!
"A modern Internet website with a scroll bar that isn't broken."
I'll see myself out now.
Preach!
How is it that my FF preference (something like 'Always show scrollbars') is regularly just ignored.
I want my scrollbars! And I wat `em fat and in my UI's color scheme!
And why the HELL would a web designer think this is a good idea. Maybe their beret and hipster pants are too tight?
Thanks for being here for my rant. I owe you one :-)
What I see happening soon is a deeper transition from playful whimsy to real applications with contractual obligations.
Right now it's an unencumbered exploration with only a few requirements. The general public isn't too upset even if it blatantly feeds on copyrighted data [1] or attacks wikipedia with its AI crawlers [2].
The end state once legislation has had a chance to catch is breath looks more like Apple being forced to implement USB type C.
[1] https://arstechnica.com/tech-policy/2024/02/why-the-new-york...
[2] https://arstechnica.com/information-technology/2025/04/ai-bo...
I don't see this as being better or worse than all the reboots, remakes, and pointless sequels for movies that make the bulk (in financing) of Hollywood's present production: these also make the same "assumptions" of how some idea should look like. In fact, I think they make it even worst.
Give me "An image of an archeologist adventurer who wears a hat and uses a bullwhip, as if he was defeated by the righteous fight against patriarchy by a strong, independent woman, that is better at everything than he is": Sure, here is Indiana Jones and the Dial of Destiny for you.
In fact, when generative video evolve enough, it will usher an era of creativity where people that previously were kept out of the little self pleasing circle of Hollywood, will be able to come up with their own idea for a script and make a decent movie out of it.
Not that I believe AI will be able to display properly the range of emotions of a truly great actor, but, for 99% of movies, it will be just fine. A lot of good movies rely mostly on the script anyway.
Guardrails seem to be ass ..... that or certain IP is already licensed
>AI is supposed to be able to [...] make extremely labor intensive things much easier
... as showcased by the cited examples?
More so, this derivative work would otherwise be unreachable for regular folk with no artistic talent (maybe for lack of time to develop it), but who may aspire to do such creative work nevertheless. Why is that a bad thing? Sure, simple posts on social media don't have much work or creativity put into them, but are enjoyable nevertheless, and the technology _can_ be used in creative ways - e.g. Stable Diffusion has been used to turn original stories drawn with stick figures into stylized children's books.
The author argues against this usage for "stealing" the original work, but how does posting a stylized story on social media "steal" anything? The author doesn't present any "pirated" copies of movies being sold in place of the originals, nor negative impact on sales figures. In the case of the trending Studio Ghibli, I wouldn't be surprise to see a positive impact!
As for the "soulless 2025 fax version of the thing", I think it takes a very negative mindset to see it this way. What I've seen shared on social media has been nothing but fun examples, people playing around with what for them is a new use of technology, using it on pictures of fond memories, etc.
I'm inclined to agree with the argument made by Boldrin and Levine:
>”Intellectual property” has come to mean not only the right to own and sell ideas, but also the right to regulate their use. This creates a socially inefficient monopoly, and what is commonly called intellectual property might be better called “intellectual monopoly.”
>When you buy a potato you can eat it, throw it away, plant it or make it into a sculpture. Current law allows producers of a CDs and books to take this freedom away from you. When you buy a potato you can use the “idea” of a potato embodied in it to make better potatoes or to invent french fries. Current law allows producers of computer software or medical drugs to take this freedom away from you. It is against this distorted extension of intellectual property rights that we argue.
https://www.researchgate.net/publication/4980956_The_Case_Ag...
> It’s stealing, but also, admittedly, really cool. Does the growth of AI have to bring with it the tacit or even explicit encouragement of intellectual theft?
You answered your own question by explicitly encouraging it.
As a consumer, is it possible for me to opt out of seeing IP protected imagery?
I would be absolutely fine with not having pokemon, mickey mouse etc shoved down my .. eyeballs.
I know this is a ridiculous point. But I think I'm getting at something here - it ought not to be a one-way street - where IP owners/corporations etc endlessly push their nonsense at me - but then, if I respond in a personal way to that nonsense I am guilty of some sort of theft.
It is a perfect natural response to engage with what I experience - but if I cannot respond as I like to that experience because of artificial constructs in law - perhaps it ought to be possible (legally) to avoid that IP protected immersion in the first place. Perhaps this would also be technologically possible now.
Won't someone think of the consumers?
So, you would like an AI that can reasonably answer questions about various aspects of human culture, or at least to take said culture into account? At the same time, you want it to not use the most obvious, most relevant examples from the culture, because they are somehow verboten, and a less relevant, completely made-up character should be presented instead? Come on, computes have logic. If you ask the machine to produce a picture of a man who was sent to humans to save them, then was executed, and after that came back alive, who do you think the machine should show you, for Christ's sake?
Picture-based industries had a mild shock when "Photoshop" kind of software became widely available. Remixing visuals in ways that the copyright holders won't approve of became equally widespread. Maybe that added some grey hairs to the heads of some industry execs, the sky has not fallen. I suppose the same is going to happen this time around.
I need a Ghibli version of that terminator now.
If OpenAI is ok with this then they should be ok with sharing their code and models so that others can profit from their work (without giving anything back to OpenAI).
Interesting how the longer conversations here go into the familiar territory of whether copyright should exist. Meanwhile, the salient aspect is that these AI image generators were trained on copyrighted material. The typical hacker news discussion feels very different when talking about code generation. Yet, when it is images - then we question whether copyright should exist?
Do you do ChatGPT illustration request-type headlines now?
Soon, the only thing that will make "Ready Player One" a fiction is that, in the end of the movie, they agree to shut the thing down once in a while.
That will never happen under Silicon Valley's watch.
That's a really blatant lack of creativity. Even if obviously all of those prompts would make anyone think of exactly those iconic characters (except possibly a different version of James Bond or Lara Croft), any human artist could easily produce a totally different character that still fits the prompt perfectly. This AI can't, because it really is just a theft machine.
To be honest, I wouldn't mind if AI that just reproduces existing images like that would just be banned. Keep working on it until you've got something that can actually produce something new.
The prompt is what's lacking creativity. The AI knows it, and produced the obvious popular result for a bland prompt. Want something more unique? Write better prompts.
> Still, the near perfect mimicry is an uncomfortable reminder that AI is getting better at copying and closer to
I completely disagree. It's not getting "better." It always just copied. That's all it /can/ do. How anyone expected novel outputs from this technology is beyond me.
It's highly noticeable if you do a minimal analysis, but all modern "AI" tools are just copyright thiefs. They're just there to whitewash away liability from blatantly stealing someone else's content.
> It's not getting "better." It always just copied. That's all it /can/ do
That's true of all the best artists ever.
> They're just there to whitewash away liability from blatantly stealing someone else's content.
That's because that's not a thing. Ownership of "content" is a legal fiction invented to give states more control over creativity. Nobody who copies bytes which represent my music is a "thief". To be a thief, they'd need to, you know, come to my house and steal something.
When someone copies or remixes my music, I'm often not even aware that it has occurred. It's hard to imagine how that can be cast as genuine theft.
are we creating a life with AI or are we creating a slave?
Paywall?
Well, property is theft, as we all know.
Unlike what they name in physics, laws in juristic field are not a given by the cosmos. That's all human fantasy, and generally not enacted by the most altruistic and benevolent wills.
Call me back when we have no more humble humans dying from cold, hunger and war, maybe I'll have some extra compassion to spend on soulless megacorps which pretend they can own the part of our brain into which they inject their propaganda and control what we are permitted to do starting from that.
> copywritten
This syntactic mistake is driving me nuts in the article is driving me nuts. It's a fundamental misunderstanding of what the the word means.
It's "copyright". It's about the rights to copy. The past participle is "copyrighted", someone has already claimed the rights to copy it, so you can't also have those rights. The rights belong to someone else. If you remember that copyright is about rights that are exclusive to someone and that you don't have them, then you wouldn't make such a syntax error as "copywritten".
The homophone "copywriting" is about take a sales copy (that is, an ad) and writing it. It's about creating an advertisement. Copywritten means you've gone around to creating an ad. As an eggcorn (a fake etymology created after the fact in order to explain the misspelling or syntax error), I assume it's understood as copying anything, and then writing it down, a sort of eggcornical synonym of "copy-paste".
I wish I wrote an article like this two days ago, when on a whim I asked ChatGPT to create something similar to a regular meme I just saw on Facebook:
Me: Can you make me a meme image with Julian Bashir in a shuttlecraft looking up as if looking through the image to see what's above it, and the caption at the top of the image says, "Wait, what?".
ChatGPT: proceeds to generate a near-perfect reproduction of the character played by Alexander Siddig, with the final image looking almost indistinguishable from a screencap of a DS9 episode
In a stroke of meta-irony, my immediate reaction was exactly the same as portrayed by the just generated image. Wait, WHAT?
Long story short, I showed this around, had my brother asking if I'm not pulling his leg (I only now realized that this was Tuesday, April 1st!), so I proceeded to generate some more examples, which I won't describe since (as I also just now realized) ChatGPT finally lets you share chats with images, so you can all see the whole session here: https://chatgpt.com/share/67ef8a84-0cd0-8012-82bd-7bbba741bb....
My conclusion: oops, OpenAI relaxed safeguards so you can reproduce likeness of real people if you name a character they played on a live-action production. Surely that wasn't intended, because you're not supposed to reproduce likeness of real people?
My brother: proceeds to generate memes involving Donald Trump, Elon Musk, Gul Dukat and Weyoun.
Me: gpt4o-bashir-wait-what.jpg
I missed the window to farm some Internet karma on this, but I'm still surprised that OpenAI lets the model generate likeness of real politicians and prominent figures, and that this wasn't yet a front-page story on worldwide news as far as I know.
EDIT:
That's still only the second most impressive thing I find about this recent update. The most impressive for me is that, out of all image generation models I tested, including all kinds of Stable Diffusion checkpoints and extra LoRAs, this is the first one that can draw a passable LCARS interface if you ask for it.
I mean, drawing real people is something you have to suppress in those models; but https://chatgpt.com/share/67ef8edb-73dc-8012-bd20-93cffba99f... is something no other model/service could do before. Note: it's not just reproducing the rough style (which every other model I tested plain refused to) - it does it near-perfectly, while also designing a half-decent interface for the task. I've since run some more txt2img and img2img tests; it does both style and functional design like nothing else before.
Copyrights and Trademarks were already a mess and now we have AI built on top of that mess.
The whole thing should be public domained and we just start fresh. /s
I just think it’s stealing, and honestly not that cool once you’ve seen it once or twice, and given the ramifications for humanity.
I know of no other type of theft that results in more of something existing in the world. Stealing deprives someone of something; copying data (from training an AI all the way to pedestrian "YoU wOuLdN'T DoWnLoAd a cAr" early-aughts file-sharing) decreases scarcity, it doesn't increase it.
The majority of human creation is like this. All language. Process. Even the US anthem was “stolen”