The problem is that the very capabilities that let a game have “way more of something than it could otherwise have” (say, thousands of unique voices reading context-specific runtime generated text) can be used to reduce the need for workers (so one can just pretty much generate all speech in game by paying a bunch of random people of the street for to come over and read text for 1h and then just clone their voices and used that to generate all in-game speech - the quality way less than pre-prepared lines read by a trained voice actor, but the cost will be a tiny fraction of it).
AI can helps us do things which in practice would otherwise be impossible but many (maybe most) companies are just using it to cut manpower costs even though it delivers inferior results than than trained professionals.
The problem is that the very capabilities that let a game have “way more of something than it could otherwise have” (say, thousands of unique voices reading context-specific runtime generated text) can be used to reduce the need for workers (so one can just pretty much generate all speech in game by paying a bunch of random people of the street for to come over and read text for 1h and then just clone their voices and used that to generate all in-game speech - the quality way less than pre-prepared lines read by a trained voice actor, but the cost will be a tiny fraction of it).
AI can helps us do things which in practice would otherwise be impossible but many (maybe most) companies are just using it to cut manpower costs even though it delivers inferior results than than trained professionals.