DeepMind’s AI teaches itself to play chess in four hours, beats top chess software

2 min read
29

chess

If you listen to the ramblings of tech wizard billionaire Elon Musk, then you’ll know that he believes that Artificial Intelligence will become the thing that dooms humanity. He believes that (if we’re not already living in a simulation), that AI will become sufficiently advanced to gain sentience, and enslave people. Terminator and its sequels aren’t just fun movies about killer robots, they’re a look at our future. He is, despite being brilliant, also mad, right?

Maybe. While Artificial Intelligence is fascinating, incredible and all sorts of other shiny superlatives, it is also terrifying. If you’ll recall, Google’s DeepMind created an AI that bested the world’s best Go players. It’s an incredibly complex Chinese board game that takes years to master. The AI learned how to play it in just 8 hours.

Now, a descendent of that AI, called AlphaZero – has beaten the world’s top chess program, Stockfish. After just four hours of teaching itself to play the game. What makes it really interesting is that the AI wasn’t designed to play chess, nor was it fed endless chess manuals and strategies. In what’s called “reinforcement learning,” the AI was given the game’s basic rules, and played against itself at an accelerated pace for four hours, devising its own strategies.

Out of 100 games, AlphaZero obliterated the world’s highest-rated chess engine – with 28 wins, 72 draws and zero losses. It’s an incredible achievement not just for the chess bit of it, but for the future of AI. Even grandmaster Gary Kasparov is impressed.

“We have always assumed that chess required too much empirical knowledge for a machine to play so well from scratch, with no human knowledge added at all,” Kasparov said. “Of course I’ll be fascinated to see what we can learn about chess from AlphaZero, since that is the great promise of machine learning in general—machines figuring out rules that humans cannot detect. But obviously the implications are wonderful far beyond chess and other games. The ability of a machine to replicate and surpass centuries of human knowledge in complex closed systems is a world-changing tool.”

It took a machine four hours to become a better chess player than all of humanity. By teaching itself to play. I think Musk may be on to something.

Last Updated: December 7, 2017

Geoffrey Tim

Editor. I'm old, grumpy and more than just a little cynical. One day, I found myself in possession of a NES, and a copy of Super Mario Bros 3. It was that game that made me realise that games were more than just toys to idly while away time - they were capable of being masterpieces. I'm here now, looking for more of those masterpieces.

Check Also

Google has mandated that new, popular Android devices offer at least two years of security updates

I’m a big fan of Android devices, but there’s one niggling problem with the operating syst…