Browse
Category
DiscussionsOff Topic

AlphaGo vs Lee Sedol (4-1)

Closed
whitywind

So AlphaGo is already victorious over Lee in a board game called Go. I guess many of you will wonder what is Go since it's almost solely played in East Asia. I'm not a good explanatory person, so here's a somewhat detailed wikipedia article of Go XD. The below is about the match between AlphaGo and Lee. If you're not so willing to read them, you can go to the bottom to enter a GA.

Let me start with that... Go is a kind of board game with perfect information.
The perfect information means that a player can see all the information needed to decide the next move in the game. Chess might be a familiar example for English speakers, as well as Checkers (Draughts). As many can guess, it's possible for AI to compute all the possible moves based on its strategy to the win since there should be no random guess, if AI has an infinite computing power. While IBM Deep Blue beat the World Chess Champion in 1997, and Checkers became solved in 2007 proving that AI can't lose any single game, Go has been the hardest board game for AI to analyze since the number of possible moves in a game is enormous. It's believed it'll be decades later when AI can actually beat a professional Go player. Well.. it was.

In October 2015, Google-made AI called AlphaGo beat the European Champion Fan Hui by 5-0. It was just a beginning. Lee Sedol, who is said to be one of the legendary professional Go players, decided to take up the challenge from AlphaGo. The match was best-of-five. Many thought Lee would win it with ease since it should be too early for AI to compete with professionals. It is a well known fact that Go AI is rather weak in general than those of other board games, because of its computational cost. Then AlphaGo won three in a row against Lee.

It upset those who previously said it should be an easy game for Lee. Actually it was a remarkable point in the history of gaming AI, revealing how far AI can go and will be... Some people mourned that there would be no position for humans on the ladder of Go and any board games in the near future. It's not that big deal though. Chess is still prevalent all over the world.

The interesting part is that, after three losses in a row that decided his overall loss, Lee got his first win in the fourth game. He studied his previous games to gather every clue to beat AlphaGo. He showed his incredible outcome on move 78 in game four, which totally exceeded AlphaGo's computing ability. AlphaGo "reported" to the dev that it made a mistake on move 79, noticed it on move 87. I'm impressed to see how Lee reacted when he was confronted by such an unknown monster, what human can do even if he is doomed.

The last game of the match will take place on March 15th, starting at 13:00 Korean local time (04:00 GMT).
ENDED AlphaGo won. Good Game.

For more information
AlphaGo versus Lee Sedol (Wikipedia)
Google DeepMind AlphaGo

Sorry for my poor English since I'm not a native speaker. (I'm trying to improve it)
If you're looking for something, here's a little GA. (hopefully relevant)

8 years ago*

Comment has been collapsed.

2 Comments
plugz

I'm looking forward to the last match !

AlphaGo didn't need to report of its mistake when you just have to watch Lee's face when he saw the move, even Fan Hui's look gave it away (Fan Hui plays alphago's move during this event) :)

I think any pro could see that it was a bad move, so obvious, Lee Sedol beeing the incredible player that he is took advantage of it but Lee already was playing an incredible match.

What was surprising to me is that Lee Sedol clearly understood the way he should manage his time, and played a much stronger first part of the game, so strong that it lead to alphago's weird move. Alphago is already an incredible AI but it is even much stronger in close combat (as were older AI's too) and in end game strategy, Lee realised that to win he needed to push alphago earlier than what he did in match 1 to 3, and that clearly showed in his approach of this game where he pushed alphago's capabilities to their (actual) limit way earlier.

The last match will hopefully be as incredible as the other ones !

8 years ago
Permalink

Comment has been collapsed.
whitywind

Yeah it'll be some peak point where we can enjoy the match of AI vs human XD
Lee said he learned something from his lost games (as well as early push strategy you said) so he might show us again, or AlphaGo simply overpower, or not...

8 years ago
Permalink

Comment has been collapsed.

Closed 8 years ago by whitywind.