Comment by jasomill
1 day ago
Not going to be viable for entire genres of games because latency.
AIs can't open source games, they can only assist in pirating them.
But pirates have historically done a pretty good job without AI.
Existing trends towards games with required online components, subscription-based content and microtransactions will no doubt continue, of course.
Not streaming, other than as an extra-cost option, because the margins on subscriptions and microtransactions are bound to be orders of magnitude higher while providing more value to the majority of players.
The reality is that proprietary HW from past console generations can be significantly improved and made less expensive via modern tech advancements to provide a superior gaming experience without load times or latency. A $200 modernized PS2 or $100 modernized SNES will dominate gaming long before streaming does.
> AIs can't open source games, they can only assist in pirating them.
AIs should be able to use the same clean room reverse engineering techniques humans use in order to produce a legally clean open source game engine that can use the original game's copyrighted assets.
... A clean room, means that you aren't reverse engineering. You're producing the same result, with zero awareness of the process, or it isn't a clean room.
LLMs are created through mass ingestion of data, and so, by definition, cannot be a clean room. They are trained on the things that they're tasked with copying.
How could the possibly have been trained on the proprietary source code of games made many years ago?
1 reply →
How is Claude going to be trained on GTA5'S game engine source code if it's closed source?
3 replies →
[flagged]
You make agent A decompile the code and describe the behavior of each function in a spec. Then agent B with no access to the internet or the original code implements it.
This is legally safe because the spec only described interfaces and had no copyrighted code from the original app. APIs are fair use under Google v Oracle and code that accidentally ends up the same falls under merger doctrine.
They can maybe try and get you on patents but that's easy to avoid by doing it in Europe. Trade secrets don't apply because you reverse engineered the code. The only thing left is good old section 1201.
Of course it's different in practice because these companies have huge legal teams and know where the corpo friendly or tech illiterate judges work. But that's a separate question from is it legal or not.
1 reply →