AlphaGo, the artificial-intelligence program developed by DeepMind, lost Go match number four against South Korean Go grandmaster Lee Sedol.
The message "AlphaGo resigns" made Lee's day on Sunday, when the clash between man and machine learning took place in Seoul. Lee snatched the victory after AlphaGo won the first three matches.
"I couldn't be happier today ... this victory is priceless. I wouldn't trade it for the world," Lee said.
Experts seem to agree that the upcoming fifth match between Lee and AlphaGo will be a toe-to-toe battle between intuition, experience and AI.
"It seems Lee Sedol can now read AlphaGo better and has a better understanding of how AlphaGo moves," says 9-dan Korean commentator, Song Taegon.
Go is probably the most complex strategy board game in the world, which requires both calculation and intuition. The fact that it has a quasi-infinite number of moves makes it more challenging, even for computers that use brute-force number crunching to come up with the next moves.
The team behind AlphaGo coded it to replicate human intuition, and gifted it with a deep-learning mechanism. AlphaGo self-improves constantly by playing against itself.
Demis Hassabis, the leader of the UK-based team that worked long hours to deliver AlphaGo, sent his regards to Lee for the victory. He also pointed out that the machine's lost game is of immense importance for improving the software. By losing, the team learns a lot about the weaknesses of the AI.
"This is why we came here: to test AlphaGo to its limits," Hassabis said.
In South Korea, the Go game is so popular that it has a strong fan base and two dedicated television channels. The confrontation between man and AI was welcomed with much enthusiasm in South Korea, where commentators went so far as to call AlphaGo's moves as "beautiful."
After winning the fourth match, Lee noted that AlphaGo seems to be caught off-guard by unorthodox moves.
"It might be a bug. I don't know if I can call it that," Lee points out.
Go differs in many ways from chess, and the first move is one such example. In Go, the black makes the first move and Lee noticed that Alphabet's program plays weaker when it uses the black counters.
In the match last Sunday, Lee played using white counters. He says that in order to test his assumption about AlphaGo's weakness, he aims to play the final match using black counters.
Experts agree that Lee secured the victory because of a sharp attack in the middle of the board. Specifically, move 78 "was really brilliant" and opened a straight path to winning the match.
Just as a reminder, the Go board features 19 x 19 lines that form a grid and players place counters on each intersection in an attempt to secure territory.
Commentators noticed that as AlphaGo was nearing defeat, its moves started to resemble those of humans that find themselves in a desperate situation.
Lee Sedol wins game 4!!! Congratulations! He was too good for us today and pressured #AlphaGo into a mistake that it couldn’t recover from
— Demis Hassabis (@demishassabis) March 13, 2016
Check out the summary of the legendary match.
