Get the latest tech news
Watching o3 model sweat over a Paul Morphy mate-in-2
A breakdown of how an AI model attempts to solve a complex chess puzzle, showcasing its human-like reasoning, problem-solving attempts, and eventual reliance on external information.
When I gave 03 a tough chess puzzle, it behaved almost like a human: thinking, doubting, retrying, and finally googling the answer. “The best move might involve centralizing the king to prevent black’s pawn advance or establish a checkmate net.” Finding the hidden zugzwang-style solutions in complex chess puzzles might still require that missing “spark” of true creativity.
Or read this on Hacker News