I think without getting into the whole AI is good/bad/terrible/amazing debate, can we just appreciate that it would have taken some developers a lot of time and coffee to make this happen without an LLM.
Another example recently is Ghost Recon 2; which originally came out on Xbox has now been ported to the PC.
> Games will in the future be streamed from the cloud
Stadia is already dead. Do you think someone else will pick that up?
To be honest I don't see the connection to LLMs with that particular problem (original assets for download). This could have been done without LLMs in the past. It was always illegal, it is now illegal and therefore it will be prosecuted if there's someone interested in it.
> The trajectory is clear I'm afraid. Games will in the future be streamed from the cloud, so that there are neither assets nor logic to steal.
That is a good point but could it be also for the following reason; cloud streaming will be "encouraged" as hardware prices keep high and will encourage a monopoly on data centres to protect those hardware asset values?
Companies fought against piracy since 80s yet here we are. Denuvo ring-0 backdoors are what, 10 years old now?
Cloud can be beaten too - once you can decompile engine, what stops you from simply fetching all the assets with 1 legal copy from cloud, and repackage game with all of them, with no internet required and piracy checks removed? Its a race and no side is considering dropping their approach.
Those are not applicable to wide audience yet, people have connection drops, lags and overall quality is lower. Also not everybody willing to buy game is on 1gbit fiber to have smooth gaming experience. One day in the future, maybe.
No? Pretty sure AI can still reconstruct the game from video stream.
Also aren't the AI models themselves copyright violations? On a much larger scale than anything they produce? It is now proven (they admitted in court, so this can be submitted in other cases as evidence and can't be disputed) that OpenAI, Anthropic and Facebook/Meta all trained their models on illegally downloaded copyrighted material, and they did so without a license to even read that. They also did that before the AI training copyright exception came into force, so even that doesn't cover it.
How reasonable is it for courts to allow the AI companies to do that, then have problems with what is produced with the models?
[dead]
in these days of the security apocalypse (almost Internet Explorer two thousand one level) should we be more hesitant to open unvetted Eckernews URLs?
This, again, is just piracy.
Stolen assets with no compensation being rendered by code built by no one.
We had piracy before LLMs. It was more honest.
Piracy was better user experience, according to Gabe. Something running in the browser is even better than that.
I think without getting into the whole AI is good/bad/terrible/amazing debate, can we just appreciate that it would have taken some developers a lot of time and coffee to make this happen without an LLM.
Another example recently is Ghost Recon 2; which originally came out on Xbox has now been ported to the PC.
https://www.ghostrecon.net/forums/index.php?/topic/75174-gho...
I never got a chance to play it as I was mainly a PC gamer, so is that a bad thing to now have that opportunity to play it 23 years later on my PC?
It is, because the downstream consequences of this will be that the industry will do _everything_ to stop this from happening.
The idea of reimplementing the _engine_ is good. Handing out the _assets_ however is not.
The trajectory is clear I'm afraid. Games will in the future be streamed from the cloud, so that there are neither assets nor logic to steal.
Previously, before LLMs, the people doing this stuff had enough friction to consider their actions. Now they don't.
> Games will in the future be streamed from the cloud
Stadia is already dead. Do you think someone else will pick that up?
To be honest I don't see the connection to LLMs with that particular problem (original assets for download). This could have been done without LLMs in the past. It was always illegal, it is now illegal and therefore it will be prosecuted if there's someone interested in it.
> The trajectory is clear I'm afraid. Games will in the future be streamed from the cloud, so that there are neither assets nor logic to steal.
That is a good point but could it be also for the following reason; cloud streaming will be "encouraged" as hardware prices keep high and will encourage a monopoly on data centres to protect those hardware asset values?
Yes, stuff rarely is monocausal.
But just because there already is fuel, doesn't mean that one has to add even more to the fire
Companies fought against piracy since 80s yet here we are. Denuvo ring-0 backdoors are what, 10 years old now?
Cloud can be beaten too - once you can decompile engine, what stops you from simply fetching all the assets with 1 legal copy from cloud, and repackage game with all of them, with no internet required and piracy checks removed? Its a race and no side is considering dropping their approach.
> Cloud can be beaten
I think they meant the kind where you only get a video stream, can't do anything there.
Those are not applicable to wide audience yet, people have connection drops, lags and overall quality is lower. Also not everybody willing to buy game is on 1gbit fiber to have smooth gaming experience. One day in the future, maybe.
No? Pretty sure AI can still reconstruct the game from video stream.
Also aren't the AI models themselves copyright violations? On a much larger scale than anything they produce? It is now proven (they admitted in court, so this can be submitted in other cases as evidence and can't be disputed) that OpenAI, Anthropic and Facebook/Meta all trained their models on illegally downloaded copyrighted material, and they did so without a license to even read that. They also did that before the AI training copyright exception came into force, so even that doesn't cover it.
How reasonable is it for courts to allow the AI companies to do that, then have problems with what is produced with the models?