The impact of ML on the games industry belongs here. The discussion about the impact on society exists in a forum that no longer exists. That discussion can rage elsewhere.
It is inevitable that tiny inputs might be creating hugely detailed 3D interactive worlds in the future that approach perfection. There is no limit in where technology can and will go.
There are absolutely limits here. Even on a philosophical level, the idea that AI will understand what we find 'good' or 'perfect' about a game is going to be so insanely difficult for it. There is far, far more to creating a good game, or even game world, than just having it look good and real. Tons of people are creating utterly stunning UE5 environments for fun, but that doesn't mean they'd actually work in a real game setting.
Design intent beyond visual factors is gonna be extremely hard for AI to crack in any kind of truly useful manner. It's something that humans are simply gonna be better at for quite a long time, because in the end, AI is not 'creative'. AI cannot be inspired. AI cannot imagine what does not already exist. AI does not understand the very specific limitations and rules and balancing factors that are super important to having a given game be remotely functional(all these things largely need to be tailored per game and aren't just constants for all), let alone genuinely good. At least not the kind of generative machine learning intelligence we're talking about today as 'AI'. And even if it were potentially capable of it, it being able to create an end-to-end coherent and entirely playable game is just so insanely far away from what it's doing today. If you want to theorize about what it could maybe do in the more distant future, go ahead, but just know such 'far future' predictions almost always age poorly because things never go how we dream they will and technology and its applications dont just universally progress linearly forever.