Ha, I like the site, and enjoyed playing some of the game!
So when I saw the "About Me" image, I thought Gen AI - https://cookie.engineer/about/me/teaser.jpg But no LLM used for that? (No judgment, just after seeing that image, and then you declaring no LLM use, I thought I'd ask!)
Well, I mean, I self-hosted stable diffusion to be able to generate the teaser images and the award images (for the ranks in the Game Over Dialog).
I only implemented the initial SD paper [1] back then so I have no clue whether the u-net architecture changed when it comes to the semantic mapping of the text/labels. Did that change and is Stable Diffusion 3 now an LLM, actually?
Sorry, you know this stuff way better than me. I thought "gen AI" that generated visualizations / images used similar code, but as I say Large LANGUAGE Models in my head, I realize it must be something else (at least in name!)
Again this should in no way implicate a diminishing of the work and accomplishment of building your site. Just noticing some kind of "AI" (ANI) was used to make an image. Maybe worth clarifying for pedants like me (but probably not worth it ;-)!)
> You know, for someone who critiques the "generative AI disease" on his wiki, it's a bit funny to be using generative AI here at all, hm?
> Instead of kvetching about parent's terminology
The point behind knowing how it works vs assuming how it works gives you a difference in understanding and perspective.
I know how it works, because I implemented the papers, and I started way before the current LLM hype. Models like NEAT, HyperNEAT, LSTMs, Bayesian RNNs, GANs, BERT, AutoBERT, AlphaGo are inherently useful if you understand how the model works, what it can do and what it can't. Those tools are great, if you know their purpose and applications.
Post-LLM agents that's a different problem, because a lot of people are assuming it's "AI" that magically does things, while it just hallucinates. So the dangers are higher when it comes to the unawareness of systemic issues and inherent responsibilities of using those tools.
(read also: Attention is all you need, one of the best papers on the topic, even more relevant these days).
PS: I've spent too much effort to comment on a shitposting account already. Anyways, have a great day nonetheless and a Happy New Year!
Hot-take: "Oh, here's a link that says if I click it, someone will show off their artistic front-end design theme."
"I see flat colors and a list of recent project news. Nothing at the bottom. Is it the wrong URL? Or does it not work on my phone? Let's go back and check the HN comments."
"The submission has a bunch of text which doesn't seem to answer my immediate confusion. There's a game somewhere? Was that the wrong link? Did it not pop up correctly? Did I need to give the site special permissions?"
TLDR: This random user unfamiliar with the site is not sure what they were supposed to focus-on and notice/appreciate, and wasn't able to figure it out within a reasonable period before needing to put the phone down again. (Not including the time to write this follow-up comment.)
Okay, so I think the problem here is that the "game" is easy to miss, when:
1. You have your audio off or volume too low.
2. You just don't interact with the bottom cookie prompt.
3. The prompt never shows up for some reason. Saw it a few times, but even after a lot of resetting cookies and reloading, I can't get it to show reliably.
I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
Could be that the avatar's speech is conflicting with an ongoing speech or sound track. Couldn't replicate it what you wrote yet, but gonna try to debug this a little more.
I had to put the initial things in a click listener, and then it's 3 seconds after the initial click when the cookie consent banner is shown (due to AudioContext API limitations and that they have to be a user gesture event that .resume()s the audio context instance).
> I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
The Debugger Views here show more details, and you can play the game just fine and toy around with it:
My browser auto-declines cookies so I also had the same thought. On reloading I can briefly see a pop-over with a lot of text in it before it disappears, so I guess that is the cookie game I initially missed.
Ha, I like the site, and enjoyed playing some of the game!
So when I saw the "About Me" image, I thought Gen AI - https://cookie.engineer/about/me/teaser.jpg But no LLM used for that? (No judgment, just after seeing that image, and then you declaring no LLM use, I thought I'd ask!)
> But no LLM used for that?
Well, I mean, I self-hosted stable diffusion to be able to generate the teaser images and the award images (for the ranks in the Game Over Dialog).
I only implemented the initial SD paper [1] back then so I have no clue whether the u-net architecture changed when it comes to the semantic mapping of the text/labels. Did that change and is Stable Diffusion 3 now an LLM, actually?
[1] https://arxiv.org/pdf/2112.10752
Sorry, you know this stuff way better than me. I thought "gen AI" that generated visualizations / images used similar code, but as I say Large LANGUAGE Models in my head, I realize it must be something else (at least in name!)
Again this should in no way implicate a diminishing of the work and accomplishment of building your site. Just noticing some kind of "AI" (ANI) was used to make an image. Maybe worth clarifying for pedants like me (but probably not worth it ;-)!)
You know, for someone who critiques the "generative AI disease" on his wiki, it's a bit funny to be using generative AI here at all, hm?
Instead of kvetching about parent's terminology when you've proven you clearly know what he meant, I suggest the disclaimer:
"No LLMs were used in the making of this website and its content, but self-hosted latent diffusion models were."
> You know, for someone who critiques the "generative AI disease" on his wiki, it's a bit funny to be using generative AI here at all, hm?
> Instead of kvetching about parent's terminology
The point behind knowing how it works vs assuming how it works gives you a difference in understanding and perspective.
I know how it works, because I implemented the papers, and I started way before the current LLM hype. Models like NEAT, HyperNEAT, LSTMs, Bayesian RNNs, GANs, BERT, AutoBERT, AlphaGo are inherently useful if you understand how the model works, what it can do and what it can't. Those tools are great, if you know their purpose and applications.
Post-LLM agents that's a different problem, because a lot of people are assuming it's "AI" that magically does things, while it just hallucinates. So the dangers are higher when it comes to the unawareness of systemic issues and inherent responsibilities of using those tools.
(read also: Attention is all you need, one of the best papers on the topic, even more relevant these days).
PS: I've spent too much effort to comment on a shitposting account already. Anyways, have a great day nonetheless and a Happy New Year!
This feels like moving the goal posts.
The site uses cookies, so needs a cookie warning. (monocle emoji)
Exactly, that's what the game is about!
Oh there is one, it just appears after a delay.
Hot-take: "Oh, here's a link that says if I click it, someone will show off their artistic front-end design theme."
"I see flat colors and a list of recent project news. Nothing at the bottom. Is it the wrong URL? Or does it not work on my phone? Let's go back and check the HN comments."
"The submission has a bunch of text which doesn't seem to answer my immediate confusion. There's a game somewhere? Was that the wrong link? Did it not pop up correctly? Did I need to give the site special permissions?"
TLDR: This random user unfamiliar with the site is not sure what they were supposed to focus-on and notice/appreciate, and wasn't able to figure it out within a reasonable period before needing to put the phone down again. (Not including the time to write this follow-up comment.)
Okay, so I think the problem here is that the "game" is easy to miss, when:
1. You have your audio off or volume too low.
2. You just don't interact with the bottom cookie prompt.
3. The prompt never shows up for some reason. Saw it a few times, but even after a lot of resetting cookies and reloading, I can't get it to show reliably.
I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
Could be that the avatar's speech is conflicting with an ongoing speech or sound track. Couldn't replicate it what you wrote yet, but gonna try to debug this a little more.
I had to put the initial things in a click listener, and then it's 3 seconds after the initial click when the cookie consent banner is shown (due to AudioContext API limitations and that they have to be a user gesture event that .resume()s the audio context instance).
> I suggest a special URL parameter so that you can link people to a slightly deeper step in the process where It's obvious what they should be interacting with.
The Debugger Views here show more details, and you can play the game just fine and toy around with it:
https://cookie.engineer/design/consent/index.html
My browser auto-declines cookies so I also had the same thought. On reloading I can briefly see a pop-over with a lot of text in it before it disappears, so I guess that is the cookie game I initially missed.
sneed
lol, this guy is 5 years too late to the doxxing party.
Lol, amazing! Love the voice.