Raw LLM Responses

​​@Yku30 Eeeeh, I wouldn't really consider chess AI to be the "pinnacle of 'smart' AI". Not to diminish Chess as a game or Chess players, but Chess is a relatively simple turn based game with very straightforward rules and a very limited amount of strategically viable possible moves or even possible moves at all compared to something much more dynamic like most modern strategy videogames nowadays. Not to mention it is a game that has been around for many centuries; More than enough time for people to make well defined and restricted "metas" that stand the test of time for you to base an AI algorithm on. Training an AI to play chess was an amazing achievement back in the day, but try to train an AI to competitively play something like Starcraft 2, Hearts of Iron 3 or 4, The Total War: Warhammer Trilogy or pretty much any other Strategy game for that matter, with all their dynamism, different possible playstyles, being much, much more assymetrical than Chess with all their different factions, units, their roles, army compositions and their purposes, different maps and terrain, the possibility to become creative and invent unconventional tactics. How many gazillion combinations of builds and moves that can be made and how many underlying factors there are to play a game like that and improvise on the fly. It is just too overwhelming for a deterministic AI to handle. It takes LOADS upon LOADS of investment to create an adaptive AI for a modern strategy videogame that can play on the same strategic and tactical level of decision making of that of a human without relying on pure brute force APM. Not that I think an AI capable of competing with humans in such strategy games will never happen, but it seems like it will happen later rather than sooner in our history. Just look at Alphastar and how it costed Deepmind to make an AI that still pretty does ONE strategy over and over again and relies on sudden bursts of APM microing to win.