DeepSeek R1 has generated a lot of excitement in the AI community. It is an open source model that can compete with top model from OpenAI and other major AI companies. As it turns out, it can also play chess for a bit. I used a simple prompt to get it to play a bit of chess against Stockfish. I simply asked it to respect all the rules of chess and play at its max level. As we played the game, DeepSeek provided detailed commentary of its thought process.
For this experiment, I used the Chessnut EVO. It is a gorgeous AI chessboard with Stockfish, Maia and LC0. It has an Android tablet built-in, so you don’t have to connect to a smartphone or tablet to use it. A few things happened when doing this experiment: DeepSeek R1 hit server limits after the first few moves. While it played the Sicilian Defense, it blundered after 12/13 moves and decided to break the rules anyway. It wanted to capture the pieces illegally and move the wrong ones. Seems it lost track of the position after the first few moves.
[HT]
