pull down to refresh
reply
that's way too far out to meaningfully do anything about. Are LLMs gonna get radically better? How fast? Will any chessplayer even bother beating them, will it be fun or exciting? Like, the IBM-Kasparov thing was cool but these days nobody cares.
Also, the specialized chess engined (Leela, Stockfish) are still gonna be vaaaastly better than the general-purpose LLMs, so what's the point?
reply
The point is to settle this bet in my favor. That should be enough, shouldn’t it?
reply
what, chess players? Predyx?
Kasparov was beating by IBM's computer almost thirty years ago; chess LLMs, the original neural nets, beat humans not long after that. Existing and popular general purpose LLMs like Gemini or Fable or Opus etc have gotten seriously good recently... but still just hallucinate insane (often illegal) moves on, like, move #25.
GothamChess's AI rundown battles were more fun a few years ago, when the machines could not grasp chess, but just argued back against the players: "No, of course I can put my queen there; No, you're wrong, I would like to just spawn a dead knight into the middle of the board"