Go Master Salvages a Victory, Showing Ingenuity in the Face of a Formidable AI
Lee Sedol, the 18-time world Go champion, struck a small victory for humankind over the weekend by finally defeating Google’s masterful Go-playing computer, AlphaGo. Sadly for Sedol, and for those cheering humanity’s champion, by that game the best-of-five series was already lost. However, his fightback highlights the way that human ingenuity can overcome brute-force machine learning.
The website Go Game Guru has a very interesting analysis of the fourth game in the historic match between AlphaGo and Sedol:
AlphaGo is very good at estimating its probability of winning. It appears to be able to do this even more accurately than the best human players. With the help of this skill, AlphaGo seems to be able to manage risk more precisely than humans can, and is completely happy to accept losses as long as its probability of winning remains favorable.
The Japanese have a name for this style of play, as it closely resembles the prevalent style of Japanese professionals over the previous few decades. They call it "souba" Go, which means something like "market price."
What Lee and his [team] had realized, was that they needed to completely upend the market.
In other words, Sedol did something that doesn’t involve reasoning over the position of pieces on the Go board. He considered how AlphaGo works, and sought to exploit weaknesses in its design.
AlphaGo is undoubtedly a fantastically clever piece of AI. The program combines two well-established AI techniques—machine learning and tree analysis—to master a game that previously seemed too complex and subtle for machines. So Sedol chose to pursue a strategy that made AlphaGo’s ability to judge risk less useful. And, on turn 79, he came up with an incredibly clever move to turn the game decisively in his favor.
Sadly, it’s too late for Sedol to claim the $1 million purse. But it’ll be interesting to see if Sedol can find further weaknesses in AlphaGo in the fifth and final game.
(Read more: The Verge, Go Game Guru, "Google’s AI Masters Go a Decade Earlier than Expected")
Keep Reading
Most Popular
Large language models can do jaw-dropping things. But nobody knows exactly why.
And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.
The problem with plug-in hybrids? Their drivers.
Plug-in hybrids are often sold as a transition to EVs, but new data from Europe shows we’re still underestimating the emissions they produce.
How scientists traced a mysterious covid case back to six toilets
When wastewater surveillance turns into a hunt for a single infected individual, the ethics get tricky.
Google DeepMind’s new generative model makes Super Mario–like games from scratch
Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.
Stay connected
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.