Hacker News Reader

Hacker News

Comment by lesiki | Hacker News Reader

Full thread

lesiki

Now: Human Chess: a variant where you can't play what AI would play.

Next: AI that can play Human Chess.

After: Human^2 Chess: you can't play what the AI above would play.

etc

I wonder if this creates distinctly new games at each level, or if it's just nonsense one level down.

View on HN

rinse and repeat that a few times, and the only remaining winning move will be not to play

The masterpiece of the reference ;)

Let's call it Human Tic Tac Toe

Would you like to play a nice game of Global Thermonuclear War?

Wouldn't human^2 chess be similar to regular chess? The human-chess AI is guaranteed never to play the regular-chess optimal move, so you can get a checkmate by always playing the optimal move (according to the engine). And unlike human chess, there's nothing preventing you from checkmating your opponent.

(I believe a chess engine could play human^2 chess exactly like it plays regular chess. A human couldn't because a human doesn't know what moves the chess engine would pick.)

Presumably human^2 chess would prohibit both the top engine move from human^1 chess and the top engine move from human^0 chess. That is, it's human^1 chess with the added restriction of not playing top engine moves.

The problem is that there may not always be 3 possible moves in a given position.

There may not always be 1 possible move in a given position either; how does regular chess handle that? (Presumably you'd use the same rule by default for human^n chess.)

It's called a stalemate, and results in a draw.

The lowest level might finally make the Bongcloud opening viable:

https://www.chess.com/blog/AcceleratedPog/bongcloud-opening-...

I loved the idea, but on further thought, the AI has a huge advantage in knowing what the engine move is, so they can never lose to incorrectly calling the last move and engine move.

Unless it was two different engines.

Are we reinventing GANs for chess engines here, or does it just happen to sound kinda similar?

I'd say let's go straight to SD and let AI paint the next move!

Never go in against a Sicillian!

This kind of thinking can either lead to total insanity or to the discovery of the halting problem or Cantor's diagonal argument.

Each level also has a logarithmically increasing number of "fantasy" meta games stacked on top, don't forget to take those into account: https://alexshroyer.com/posts/2022-04-30-Fantasy-Fantasy-Foo...

It stops one level in.

AI can't win at human chess, because any move that it attempts to make is by definition the top move choice of an engine, and so causes immediate defeat.

Wouldn't the perfect AI for Human^2 Chess be just the original AI you started with?

The meta is different because they know their opponent also can't play the best move which impacts what move you play (ie. intentionally hanging queen for advantage)

Reply on news.ycombinator.com