Spending 10 months building the most human-like chess bot
AI summary
Editorial policyThe article details the development of a human-like chess bot and presents a performance scorecard comparing key models.
Key points
- •Chessbook-1 has 199M parameters with a human move prediction score of 96, while Allie 2.0 has 5.6B parameters with a score of 99.
- •Maia-3 has 79M parameters with a human move prediction score of 85 and lacks think time prediction capability.
- •The models were graded on a 0-100 scale across five metrics: move prediction, strength, time scramble, think time prediction, and winner prediction.
#LLM / Model Trends#Research & Papers#Chessbook-1#Maia-3#Allie 2.0