Programming Note: August 2016 edition

I finally made myself finish analyzing my games from the Space Coast Open, and also the games from last Saturday. Some of it is actually interesting, instructive even, and some if it is tragi-comic. The more entertaining/instructive bits will be coming shortly. But I'm really proud of the coming lectures on how to become a Class B player - I think it can help a lot of people!

Now it's GO's turn.

Google has announced it has made a breakthrough in programming a computer to play GO. Up until now, programmers hadn't been able to write a decent GO program, at least not compared to human masters. But Google (and perhaps soon, Facebook) has created a program that has beaten the best human player in Europe by a 5-0 margin.

The research has implications beyond an old Chinese board game. The systems used by Facebook and Google were not preprogrammed with specific if-this-then-do-that code or explicitly told the rules. Instead, they learned to play at a very high level by themselves. These techniques can be adapted to any problem "where you have a large amount of data that you have to find insights in," Hassabis said.

One of the programmers also created a similar program for chess:

DeepMind [Google's AI research group] recently hired Matthew Lai, a London researcher who developed a system capable of playing chess at the grandmaster level. His software was able to reason in a way similar to how humans do, a more efficient method than IBM's attempt to crunch every possible outcome before making a move in the 1990s.

Here's more on Lai's prior chess program, Giraffe. Giraffe is impressive, but still not up to snuff with the best "regular" chess programs.