Archive for September, 2015

A chess computer that can think like a human?

In the post WWII period, chess has been taken as the perfect experiment in the search to make a computer ‘think’ like a human. In practice the developers quickly lost sight of that. Instead they focussed on how to make a computer beat a human, a project that involved increasing the brute force of the machine at the cost of the original goal of understanding how humans think and transferring that ability to computers.

Now an English MA student has made what looks like a breakthrough in the original plan. While bridge, as has been the case for many years now, only has any publicity in the mainstream press for cheating – and my, what a field day at the moment – chess is in the press for all sorts of reasons, this being one of them. The Daily Mail broke the story with the headline

Computer that can learn from its mistakes teaches itself to become a chess grandmaster

If you want the story complete with the Daily Mail’s righthand sexist strip on women, you can get there here.

If you want to read the MA thesis by Matthew Lai, complete with politically correct usage of gender language, on how it was done,

Giraffe: Using Deep Reinforcement Learning to Play Chess

you can find it here.

It is short and readable, I do recommend it. On can only assume, since the original idea was to make computers in general capable of human learning, that this could be a defining moment for the course of computer learning in the world in general. Exciting!

September 20, 2015 at 5:36 pm Leave a comment

September 2015
« May   Dec »

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Join 11 other followers