The Most Human Chess Engine

If you’re at all familiar with chess, you know that AI has long been dominant in the game compared to humans. Most Grandmasters and Super Grandmasters of chess now utilize Artificial Intelligence to analyze their own games to find mistakes and areas to improve. But playing and practicing against AI is very different than playing against a person, especially for lower rated players (newer to the game.)

This is because strong chess engines spot difficult combinations very easily and engines have no mating patterns or positional set-ups to rely on or to hold them back. A strong player is inclined to move toward these patterns and set-ups while an engine will be able to spot weird un-positional moves that don’t make much sense to a human player. Engines have no preference for normal or known positions, if it sees something better, it will play it. This can lead to a lot of odd looking moves. Every beginner learns to simplify the board and remove extra pieces to get to a less complicated ending. Engines do not worry about this as they never lose track or blunder pieces in the complexity. Though engines do have some issues with strong defensive positions (fortress) and that will be able to show best between a human player and engine. In time, it can be expected that engines will be able to work through these positions as well.

Introducing Maia Chess, a human-like neural chess engine designed to play the human move. Maia is an engine that learns from human games instead of self-play games, with the goal of making the most human move and not most optimal. Leela and Stockfish are other engines that try to match human moves, but they only reach 43% and 38% respectively. Maia on the other hand predicts up to 53% of player moves and as result is the most natural, human-like chess engine to date. Maia was developed using code adapted from Leela which in turn is an open-source clone of Alpha Zero, a revolutionary AI program created by DeepMind.

There are 9 trained versions of Maia, one for each Elo milestone between 1100 and 1900. Each version is only trained on human players of the same ranking, so Maia 1900 is trained only on games of 1900-rated players. Each version has learned from 12 million human games and is still learning by playing real players at Lichess, a popular online chess server.

In current work, Maia is being pushed to predict the moves a particular human player would make. By starting with a base model and training the model on a specific player’s games they are able to personalize the engine. This has given the engine the highest results and raises the accuracy to 65% for the particular player.

What does it mean?

Chess is a great way to develop and study AI. It’s popular, has well-defined rules, and it has not yet been fully solved. Many poeple know the game and AI researchers use it as a “model system” to study new ideas or techniques.
Chess players have their own playing styles, so it can be more difficult to predict their moves over the best possible move due to the astronomical amount of potential chess positions. Even the same player may make a difference move on a position that they have already seen for any various number of reasons.
Having a chess engine that plays the same way a human does at a specific Elo is a great way for players to practice and improve their own play. A normal chess engine at a specific Elo will still play like a chess engine but just make more intentional mistakes. Maia will play like a real human player and will give players a more genuine experience.

When using a particular players games for Maia to analyze and learn from, it raises the prediction rates even higher. So for more professional players that want to practice against another competitor, it will be possible for them to compile all their online games into the engine and have Maia imitate them. Allowing for a more personal training session than could be given in any other way.

Line of Research

This line of research into an AI that tries to be more human can eventually be used elsewhere as technology improves. Being able to have training/teaching software with an AI that understands which parts humans have most trouble with and what solutions have the best results to raise proficiency. For AI systems to be able to behave in more humanlike ways will allow for more people to understand how AI systems work. It can make it more acceptable for AI systems to be accepted in the future. Instead of the idea that AI systems will replace people, it will be a way for AI to be used to augment humans

Garry Kasparov, the former world chess champion, lost to the IBM’s Deep Blue over 20 years ago and he believes that AI can make us ‘more human.’ Which is definitely a thought provoking statement and would love to hear what others think about it.

11 comments

  1. Very interesting topic! One of the most difficult human traits for AI to mimic is creativity. Sure, a machine can learn every possible move of a game, but humans often don’t choose the most optimal move because they either didn’t see it or have a different strategy. I know the defense and aerospace industry has put a lot of research and development into similar human-life AI to support flight simulations. For example, many are familiar with the Miracle on the Hudson case or the movie Sully that portrayed the story. A large part of the legal case was based on an AI simulator that said the pilot made the wrong move (to not turn around). It was argued that the simulator lacked human intuition and lacked the ability to consider less optimal, yet successful, solutions. At the end of the day, AI is here to guide and assist humans; Humans are responsible for interpreting the results.

  2. I didn’t realize that chess was a way to learn and develop AI, but I can see how it can easily be translated. When it comes to ‘AI making us more human,’ I think it depends on who is behind the technology. There are some AI systems out there, like you pointed out, that are looking for the best possible outcome regardless of how ‘human’ the move might be. If that is the intention behind all AI, I don’t think that will make us more human.

  3. Nice post. We’ll cover AI in a few weeks and can get into the implications more deeply.

  4. Using chess to develop and test AI makes a lot of sense due to the complexity and logic required to win. Really like how you ended your blog with a question as well to kickoff the comments. I think the key to being human is evolving physically, emotionally and cognitively and so by that definition AI can absolutely help us become more human. I wouldn’t say that AI is key to being more human or anything but if we choose to use AI to help us learn and evolve, it could become part of what makes us human. How we use or leverage AI is up to us and I think the more we explore the better.

  5. While I was reading this post I was thinking of the Netflix show “The Queen’s Gambit,” great show if you have not watched. Coming back to AI, I looked up Maia and I saw that it observes human errors and a possible use for it could be in health care. “One way to do this is to take problems in which human doctors form diagnoses based on medical images, and to look for images on which the system predicts a high level of disagreement among them.” That sounds cool! https://www.wired.com/story/new-artificial-intelligence-mistakes-purpose-chess/

  6. I’ve been fascinated by this subject for many years. I remember in 1997 when Deep Blue, the machine built by IBM, beat a world chess champion. After Deep Blue, it became nearly impossible to win a game of chess against a computer.

    I hadn’t heard of Maia but it follows what I think will be the next big thing for artificial intelligence, connecting to humans. Maia’s ability to relate to human moves and not solely rely on its computer power is interesting. It makes me wonder if this is the first step for artificial intelligence to try and learn how we think more deeply, and relate to human thought. I look forward to seeing where this technology goes in the coming years.

  7. Chest was the game you would play while in recession growing up in Spain on a rainy day at school. A chest board brings me great memories from my childhood. Although I was never good, I founded it fun and challenging. I had fun playing against other players, but I didn’t have a chance against any computer, even on the most accessible mode. The machine will always get the best of me. What I find the most fascinating about the whole process is how machine learning could even predict my errors and erratic behavior. Always a step ahead. This blog has shone some light on my frustration as a young player.

  8. Impressive blog. I wasn’t too fond of chess as a kid, but your blog really captured my attention with respect to the current aim of automation: no longer is efficiency most useful, but the focus on replicability of human nature might be more insightful. Your blog reminded me of a research study I found regarding robots and the degree of humanoid within robots that people are willing to accept (before it begins creeping them out). While your blog focuses more on humanoid capabilities as opposed to humanoid appearances, the Maia Chess definitely taps into this research. Ultimately, I would assume the public’s tolerance of humanoid robot capabilities would be much higher than that of humanoid robot appearance.

  9. This is a really interesting topic, and I can’t wait to learn more about the implications of the technology when we study it in a few weeks. In a previous course, I remember discussing the challenges of AI, and how while specific moves in chess were able to be learned, overall strategy was a whole new aspect of AI that hadn’t been explored yet, and the key example was the video game starcraft. It seems like that’s already gone out of the window. 2 years ago, Google’s AI was able to achieve grandmaster status on the game and beating out 99.85% of all players in the game. I suppose there’s still a bit of room for growth to reach the status of Chess, whereby no human could beat a computer, but the rate of growth is extraordinary.

    https://www.nature.com/articles/d41586-019-03298-6

  10. I’m glad you touched on the human element that the machine tries to incorporate into Maia. This is what fascinates me the most about this system as that it can mimic human error and what was what I would hope would be mentioned in your blog as I reading through it. When I think of playing chess against the computer in the past I basically think of losing because it always finds a way to beat you since it’s designed to take the optimal move. It’s cool knowing that Maia has learned from human behavior through playing against humans and tries to incorporate that into the programming. I agree that it would be really useful for practice. Chances are that when you practice against a human, the human is going to trip up from time to time whereas a computer like Deep Blue would not have. A human will also go about a move different ways and may strategize the same move differently like you mentioned in your post. Deep Blue probably wouldn’t do this. So I agree Maia is likely a very valuable training tool for someone looking to improve their Chess skills

  11. Enjoyed this blog as it showed a connection between personal interests and digital transformation. Furthermore there was a clear amount of effort shown through background research and overall knowledge of the topic. I’d be interested to learn more about this topic, potential for a class presentation? Either way I’ll be downloading some chess programs later today!

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: