Instead of defeating us mercilessly at chess, artificial intelligence technology is now trying to match our level of play — mistakes and all.
New research involving computer scientists from University of Toronto, Cornell University and Microsoft is working on an AI chess engine where playing like a human is the main goal.
As Cornell University’s writeup says, computers are the undisputed chess champions of the world, as no one has defeated a chess computer in 15 years. So, this current research focuses on how humans can learn and become better.
“Current chess AIs don’t have any conception of what mistakes people typically make at a particular ability level. They will tell you all the mistakes you made — all the situations in which you failed to play with machine-like precision — but they can’t separate out what you should work on,” Ashton Anderson, assistant professor at University of Toronto said.
Known as Maia, the new chess engine is fully playable online, where you can try to beat the AI and claim victory for humankind over the machines.
“Maia has algorithmically characterized which mistakes are typical of which levels, and therefore which mistakes people should work on and which mistakes they probably shouldn’t, because they are still too difficult,” Anderson said.
Aside from generating better learning experiences, Maia is intended to be more enjoyable to play against, considering it’s designed to play the human move, which is not always the best move.
In one game against Maia 1100, I somehow eked out a stalemate despite not having played chess in years and not having the faintest clue about strategy.
The team behind Maia has got the engine to match human chess moves at a rate hovering over 50% across each skill level, which is higher than the default Stockfish AI Lichess uses.
You can check out more about how Maia was built via the research paper “Aligning Superhuman AI with Human Behaviour: Chess as a Model System”.
Sources suggest that Maia is in talks to sign on as the lead for The Queen’s Gambit season two, involving a plotline about playing chess against fellow AI opponents HAL 9000, Agent Smith, and Skynet.
Do you like what you're reading on Byteside? We're building a diverse, remotely distributed team of Australians to cover the digital world we love so much. And we need your support to thrive.
Whether a one-off donation or becoming a monthly supporter, every little helps pay the writers who are working here.
We're not in line for any free money from Facebook or Google. And we're not paywalling any content because we want everyone to be able to access what we do whether they can afford to pay or not. If you're one who can, a few dollars really does help us grow the pool and support writers to do great original work about tech, games and digital culture. Support Byteside now.