Comment

Comment on Peter Molyneux thinks generative AI is the future of games, all but guaranteeing that it won't be

I mean, if someone creates a game with all the options there and you just use AI as a replacement for a complex UI, it could kinda work. A game like scribblenauts could theorically implement an AI based stage creation option with the current tech already. The problem with that is that the AI wouldn’t be able to guarantee that the stage has a proper challenge level (or even that is possible to complete it), so it would also need to implement an AI that tries to beat the level as well and then keep iterating over the two until a proper stage is found.

In short: doable, for very niche cases and probably taking a very long time to complete a prompt (possibly hours).

source

Sort:hotnew top

ATDA@lemmy.world ⁨2⁩ ⁨months⁩ ago
Does candy crush use AI or is there just a hellish algorithm that generated thousands of levels.

source
CleoTheWizard@lemmy.world ⁨2⁩ ⁨months⁩ ago
I see AI as being more useful in things like Bethesdas radiant quest system. Theoretically an AI could generate quest and character dialog and react in unique ways to game world events. As far as game elements, machine learning is actually a pretty good way to have dynamic difficulty where the player is pushed as far as they can go and game elements are tweaked accordingly. Or the AI could even design unique quest items and names if trained right.

Plenty of applications for it but I think we’ll see it overused in some games which will lead to bland or non-cohesive elements that on the surface are fine, but don’t amount to anything unique. Like imagine cyberpunk but written by AI and it’d be mostly generic dialog with few connecting ideas. It’s not impossible for AI to get better at that though and maybe if it were only trained on other game dialog or if it gets approved by a human first, it could be incredible.

source