Gemini

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Pedro
Posts: 29
Joined: Mon Oct 26, 2020 3:05 pm
Full name: Pedro

Re: Gemini

Post by Pedro »

I posted a mate-in-two puzzle for the leading AIs to solve, and they all got it horribly wrong. They all hallucinated. Grok 3 in Think mode, Gemini 2.0 flash thinking experimental, ChatGPT in Reflect mode, Deep Seek in R1 mode, and Copilot in Think Deeper mode—all of them got it wrong and hallucinated.

However, only ChatGPT managed to get the first move of the sequence right, which is Nh5! In other words, in this puzzle, the winner was ChatGPT because it at least got the first move correct.

The command I used was this: You are a chess master. Based on that, analyze this chess position described by the FEN '1K1R3N/Bp1pp2p/1r2P2p/1p1rkp2/3R1N2/8/B2P2Q1/8 w - - 0 1' to find a mate in 2 moves for White.

The position:

Image
Top

Return to “Computer Chess Club: General Topics”