TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Show HN: Maia, a human-like neural network chess engine

10 pointsby reidmcyover 4 years ago

1 comment

reidmcyover 4 years ago
Hi everyone,<p>We&#x27;re happy to announce a research project that has been in the works for almost two years! Please meet Maia, a human-like neural network chess engine. Maia is a Leela-style framework that learns from human play instead of self-play, with the goal of making human-like moves instead of optimal moves. Maia predicts the exact moves humans play in real online games over 50% of the time. We intend Maia to power data-driven learning tools and teaching aids, as well as be a fun sparring partner to play against.<p>We trained 9 different versions on 12M Lichess games each, one for each rating level between 1100 and 1900. Each version captures human style at its targeted level, meaning that Maia 1500&#x27;s play is most similar to 1500-rated players, etc. You can play different versions of Maia yourself on Lichess: Maia 1100 (<a href="https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia1" rel="nofollow">https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia1</a>), Maia 1500 (<a href="https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia5" rel="nofollow">https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia5</a>) and Maia 190 (<a href="https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia9" rel="nofollow">https:&#x2F;&#x2F;lichess.org&#x2F;@&#x2F;maia9</a>)<p>This is an ongoing research project using chess as a model system for understanding how to design machine learning models for better human-AI interaction. For more information about the project, check out <a href="http:&#x2F;&#x2F;maiachess.com" rel="nofollow">http:&#x2F;&#x2F;maiachess.com</a>. We published a research paper (<a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2006.01855" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2006.01855</a>) and blog post (<a href="http:&#x2F;&#x2F;csslab.cs.toronto.edu&#x2F;blog&#x2F;2020&#x2F;08&#x2F;24&#x2F;maia_chess_kdd&#x2F;" rel="nofollow">http:&#x2F;&#x2F;csslab.cs.toronto.edu&#x2F;blog&#x2F;2020&#x2F;08&#x2F;24&#x2F;maia_chess_kdd&#x2F;</a>) on Maia, and the Microsoft Research blog (<a href="https:&#x2F;&#x2F;www.microsoft.com&#x2F;en-us&#x2F;research&#x2F;blog&#x2F;the-human-side-of-ai-for-chess&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.microsoft.com&#x2F;en-us&#x2F;research&#x2F;blog&#x2F;the-human-side...</a>) covered the project here. All of our code is available on our [GitHub repo](<a href="https:&#x2F;&#x2F;github.com&#x2F;CSSLab&#x2F;maia-chess" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;CSSLab&#x2F;maia-chess</a>). We are super grateful to Lichess.org for making this project possible with their open data policy.<p>In current work, we are developing Maia models that are personalized to individual players. It turns out that personalized Maia can predict a particular player&#x27;s moves up to 75% of the time. You can read a preprint about this work <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2008.10086" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2008.10086</a>.<p>We&#x27;d love to hear your feedback! You can contact us at maiachess@cs.toronto.edu or on our new Twitter account @maiachess.