You might have heard about our [cached]recent games with AlphaGo in China, at the Future of Go summit. No only did we play the legendary Ke Jie, but there were also two new and exciting formats: Team Go and Pair Go.
This match was also very exciting on the technical side because we had improved AlphaGo to the point where we ran it on [cached]a single machine in the Google Cloud - that's [cached]one tenth of the computation power compared to the distributed version we used in the last match!
Personally, I also really enjoyed the Pair Go format, both when I tried it myself and watching the game with professional players.
But first things first, let's start with the games against Ke Jie. As expected, these were some of the highest quality games of Go I had ever seen, really exciting to watch.
Others have provided much better analysis of these games than I ever could, so I'll let them stand as they are. See below for the game records, you can see full videos of all the games as well as short summary videos on the [cached]official event website.
As for Pair Go: I really enjoyed this format because even for someone like me (a weak kyu player) it makes it possible to play a game with AlphaGo at its full strength. The setup is simple: On each side there is a human player and AlphaGo, with moves played in alternation.
This makes for very interesting dynamics: As a weak amateur player, often I might not know a good response to a complicated battle. However, I can be a bit cheeky and play a move elsewhere on the board with a forced response, letting AlphaGo handle the situation on the next turn.
Similarly, in other situations AlphaGo might completely ignore my move (say a relatively small local response) and play a very big move elsewhere on the board, alerting me to an opportunity that I might have completely missed otherwise.
This makes for a very fun way to learn, and keeps play at a much higher level!
The team game was also very interesting, and it was great to see the team have so much fun during their game.