Hacker News new | ask | show | jobs
by Animats 3745 days ago
From the article: "It is estimated that, of South Korea’s three hundred and twenty pros, only around fifty are able to earn a living on tournament winnings." This isn't going to result in massive unemployment.

However, cheating with computer assistance is likely to become a problem, as it is in chess.[1] (The state of the art in computer chess is now roughly at "laptop with off the shelf program can curb-stomp human world champion.")

[1] http://en.chessbase.com/post/yet-another-case-of-cheating-in...

1 comments

What kind of configuration would you need to run the DeepMind setup? Looking at the wikipedia page if we assume the setup [1] there then it seems like at this time it would be out of the realm of feasibility at this time of setting up this software for cheating.

[1] https://en.wikipedia.org/wiki/AlphaGo#Hardware

Google's infrastructure gives it an extra boost, but even the single machine version can beat the distributed version 25% of the time.[1]

[1] https://twitter.com/demishassabis/status/708489093676568576

The nice thing about most of this deep learning stuff is that you can use a million machine hours to train your model, and almost no time to make an evaluation. So the single machine version has all of the pattern recognition given to it by the cluster, but a few ply less tree search depth.

So to me this underscores the relative importance of the deep learning model vs the tree search.

Kind of cool to know (if you're Lee) that your thinking ability is on par with 1920 CPUs and 280 GPUs!
Well, the whole point of this match has been to demonstrate that it is not...
Flip it around: it took 1920 CPUs and 280 GPUs to beat him!

I'd be ecstatic if I could keep GNU Go on 1 CPU from wiping the floor with me... :)

It wouldn't be expensive to rent that much "cloud" hardware for a few hours.
But given the strength of the distributed program, the desktop version is already professional strength because of diminishing returns.