AlphaGo

AlphaGo is the computer-Go system developed at Deepmind which took the world by storm in three phases:

  1. January 2016: it was announced that in November 2015 it had defeated Fan Hui 2p 5-0 in a secret match.
  2. March 2016: an upgraded version of AlphaGo defeated the famous Lee Sedol 9p 4-1 in a public match.
  3. January 2017: a yet stronger version of AlphaGo won 60-0 in fast games against most top pros, playing "undercover" on online Korean and Chinese Go servers (using the usernames Magister and Master).

Personally, I was deeply moved when I read in January 2016 that it defeated Fan Hui - since everyone thought that the time when a computer program can defeat a professional player is many years in the future; I was shocked to see it defeat Lee Sedol a few months later; and I was very excited to see the "new Go style" played by the the "Master" version of AlphaGo in the 60 fast online games against top Asian pros.

AlphaGo proved that Go is much more complicated that we thought, and even after perfecting this game for hundreds of year, humans are still easily defeatable by AlphaGo using very simple-looking moves.

How strong exactly is the Master version of AlphaGo? The professionals playing with it had various reactions and gave different estimations. Personally I believe it can give 2 stones handicap to top pros and still win.

You can re-watch all five games against Lee Sedol on Deepmind's Youtube channel. More game commentaries can be found on the American Go Association's Youtube channel.

An invaluable resource consists of the five commented games with Lee Sedol and three self-playing games of AlphaGo's - all can be explored either as SGF or downloaded as PDF from Deepmind's website.

Here are some interesting moves from the 60-0 series - the marked move is AlphaGo's:

For more complete and up-to-date information, see Sensei's AlphaGo page.