the key component of intelligence is “content.”
Could you explain what you mean by that? I understand it as “data is fundamental to deep learning, like compute or algorithms”, but I’m not sure I got it completely.
While I agree that data is critical to current LLMs, I’m unsure about AIs in general, as AlphaZero was trained without any data, requiring just compute, algorithm and a simulation environment of the game.
If I understand you correctly, one view is scoring (how many points I have), the other one is (how likely am I to win—the eval bar). We could have both perfectly valid views in a two player’s game
However, the eval bar for 4 players (which would be a eval vector of dimension N_player − 1) is harder to compute and wouldn’t capture effects such as alliances, or the “king maker”, a player which has lost but decides who wins. So I don’t think “optimal play” is just a matter of algorithm like the minimax can be in chess or Go.
Furthermore, in most multi-players board games, it’s almost as easy to win a point that it is to make somebody loose a point. In those games, the dynamic is different : minding your own business is a better strategy than trying to make all your opponents loose, at least until the very end of the game (where you have only one potential opponent for the win).
This is my understanding of the “big world” in board games.