reidmcy

Yes, the output is just a large vector with each dimension mapping to a move. But we'd probably do it as different "head" so have a small set of layers that are trained just to predict resignations. Both options would break the lc0 chess engine we use for stuff like the Lichess bots though.

They're trained on everything but Bullet/HyperBullet. The neural network just predicts moves and win probabilities so we don't have a way (yet) of making it concede.

We did not, we removed bullet games because they tend to be more random, and also did some filtering of the other games to remove moves made with low amounts of time remaining for the same reason. But the models don't know about different time controls right now.

Thanks for mentioning that. The graphs do use different dashes to distinguish the colour palletes, which are supposed to be colour blind friendly. So I thought we'd be OK. I'll take a look into adding more non-colour distinguishing features.

Yes, that's exactly one of our goals. In the paper we even have a section on predicting which boards lead to mistakes (in general). The results were much weaker than the move prediction, but we're still working on it and will hopefully publish a followup paper.

That's what a few machine learning people I talked too thought would happen. There are lots of examples (self driving cars being the big one) in machine learning where training on individual examples isn't enough. We were actually hoping for our models to be as strong as the humans they're trained on so we are underperforming our target in that way.

I'm the main developer for Maia chess

Hi everyone,

We'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.

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's play is most similar to 1500-rated players, etc. You can play different versions of Maia yourself on Lichess: Maia 1100 (https://lichess.org/@/maia1), Maia 1500 (https://lichess.org/@/maia5) and Maia 190 (https://lichess.org/@/maia9)

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 http://maiachess.com. We published a research paper (https://arxiv.org/abs/2006.01855) and blog post (http://csslab.cs.toronto.edu/blog/2020/08/24/maia_chess_kdd/) on Maia, and the Microsoft Research blog (https://www.microsoft.com/en-us/research/blog/the-human-side...) covered the project here. All of our code is available on our [GitHub repo](https://github.com/CSSLab/maia-chess). We are super grateful to Lichess.org for making this project possible with their open data policy.

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's moves up to 75% of the time. You can read a preprint about this work https://arxiv.org/abs/2008.10086.

We'd love to hear your feedback! You can contact us at [email protected] or on our new Twitter account @maiachess.

That doesn't help you prove they're indisputable from humans, even if you admitted to using them to cheat. You said some engines are indisputable from humans, can you give me some evidence of this?

Thanks, those do look like they are attempting to play like humans, but neither says how that claim was tested. How would you test the their ability to imitate humans?

I'm not sure if saying engines don't make long term plans is accurate, since we can set them to plan an arbitrary number of moves ahead. I think you mean that they don't have a theory of mind, and thus don't modulate their play based on their opponents.

I don't think Play Magnus is close to indistinguishable, it's a modified version of Stockfish on the backend, having the same rating doesn't mean you have the same style

Can you give an example? I haven't seen any that are indistinguishable from humans

It's working for me, just takes a couple seconds to load

SEEKING WORK - Waterloo, ON (Canada) or Remote

Just graduated University of Waterloo with a degree in Mathematical Physics.

I know: C++, python and OpenGL

I am currently working on a python library to analysis collaboration networks of scientists using IPython/Jupyter.

github: https://github.com/reidmcy email: [email protected]