The Orion's Arm Universe Project Forums
Google DeepMind Beats Go Champion - Printable Version

+- The Orion's Arm Universe Project Forums (
+-- Forum: Offtopics and Extras; Other Cool Stuff (
+--- Forum: Real Life But OA Relevant (
+--- Thread: Google DeepMind Beats Go Champion (/showthread.php?tid=2038)

Google DeepMind Beats Go Champion - PortalHunter - 01-28-2016

I'm sure many of you will find this interesting.

RE: Google DeepMind Beats Go Champion - chris0033547 - 01-28-2016

Yes, Demis Hassabis promised the world a "surprise" when it comes to Go a while ago:

Now even those games that put long term strategy above tactics are falling to AI. And I get the feeling that all of this is happening rather quickly.

RE: Google DeepMind Beats Go Champion - Bear - 01-29-2016

The wonderful thing about learning algorithms is that they can spot "patterns" so subtle that we don't even notice we've learned them ourselves, and could never express them in words.

We are *FAR* beyond attempting to score the board by any easily-expressible rule.

Go, like all turn-based games, has a self-organizing property. If you can make even a loose approximation of a "this board is better for white than that board" function, then using that board-evaluation function as the "leaf node" in a min-maxing tree one or two levels deep provides a much stronger approximation. And assuming there is any underlying pattern better than min-maxing, that stronger approximation is a pattern that a learning algorithm can learn. The learned pattern can be used as a better board evaluation function, rinse repeat.

I'm going to state confidently that there is NO turn-based game at which humans can consistently beat computers any more. Not once you've sicced a learning algorithm on refining the board-evaluation function and given it time to run.

That said, it will suck up a lot of computer power to learn iteratively better board valuations. It could take a warehouse full of machines months to do it. But once it's done, a single reasonably-powerful computer can use the learned pattern to beat any human master.

RE: Google DeepMind Beats Go Champion - Rynn - 01-29-2016

So, question for the resident expert and group at large: what aspects of human life could be classed as turn-based and be ameanable to automation by this sort of software in the near future?

RE: Google DeepMind Beats Go Champion - chris0033547 - 01-29-2016

I'm certainly no expert at this but I heard that Go is a special case of the Multi-armed bandit problem. (The Wikipedia article contains some practical applications.) Approximately 10 years ago Sylvain Gelly wrote the following paper about it:

Sylvain Gelly is one of the authors of MoGo:

The following date is important:

Quote:7 August 2008 MogoTitan vs. Kim MyungWan 8p on KGS

It was the match against the supercomputer Huygens. At that time a supercomputer needed several handicap stones in order to have a chance against a human Go master.

And now back to the present:

Kim MyungWan will comment the games of Fan Hui against AlphaGo:


A few more links:

RE: Google DeepMind Beats Go Champion - PortalHunter - 02-08-2016

[Image: reassuring.png]

RE: Google DeepMind Beats Go Champion - chris0033547 - 03-13-2016

I'm surprised that nobody mentioned the feats of the current version of AlphaGo in this thread.

There are two games left. I wonder whether Ke Jie will still play AlphaGo after this. He might have changed his mind after witnessing Lee Sedol's "annihilation" in the 3rd game. That's because Ke Jie thought that AlphaGo might not understand the concept of ko. But that assumption turned out to be wrong. The 3rd game shows that it understands ko fights.

RE: Google DeepMind Beats Go Champion - Dark - 03-13-2016

Some rather interesting commentary by Eliezer Yudkowsky.

Basically, superhuman AI performance in a particular field would not look immediately obvious. Rather than outright domination, there will be a invisible shifting of odds to the AI's favor for a while, until this advantage becomes big enough to be noticeable for the human player. And by this time it is too late for any counter-strategy to be effective as the game is already lost.

Lesser intelligences in OA universe are surely familiar with this phenomena Smile

Quote: AlphaGo is still a nice accidental illustration that when you've been placed in an adversarial relation to something smarter than you, you don't always know that you've lost, or that anything is even wrong, until the end