The game of Go has been beaten by computer

6 posts / 0 new
Last post
The game of Go has been beaten by computer

What's next? The game of love, I guess

Re: what's next

Let's wait can see the final score of the match. There are a few obvious things to follow. For the AI program to claim to beat human-beings, it must also play again other strong player, the most obvious one being Ke Jie (who has a 8-2 score against Lee Sedol). Then some strong players may team up to fight the AI.

The writing is on the wall

The 4-1 score for AlphaGo is perfect: on the one hand it is clear that the computer is super strong already, on the other hand the loss in the fourth game is a bit puzzling, not because of the move missed or misinterpreted but because of what happened next, the "immature" reaction of AlphaGo, playing a string of bad moves in the "hope" to lure Lee into some mistakes: this gives the AlphaGo team something to improve. From here there is only one direction: the level of AlphaGo and its competitors will become unreacheable for the human players. After a short while there will be no computer - human matches anymore, what would be the point, but the computers will become training devices for the human players, the top: for the amateurs there will be no fun in it, losing all the time. What I hope is that better strategies will emerge, and the bad ones, maybe still played by even top players, will become apparent along with their refutation. That will be the gain of it all.

Kind regards,

To put things in perspective:

To put things in perspective:

The Alphago team had a minimum of 15 people (visible in one of the photos taken at the end of the match) + as I understand the Nature paper they published, a minimum of 1206 CPUs x 40-thread searches x 1,000 rollout searches per second = 48 MILLION searches per second. Given that the games lasted 4 hours and the datacenters they had on tap in the 'cloud' continued searching during the opponent's moves (as also stated in the Nature paper), we're talking about 48 x 60 seconds x 60 minutes x 4 hours = 691,200 MILLION searches over a game, divided by its 140 moves (to take the last game as an example) = 4,937 MILLION searches per move. Added to this, they initially input 30 million positions from 160,000 top-level dan KGS games + 8 million positions from top-level dan Tygem games and then ran through 340 million 'training steps' before the Fan Hui games (and who knows how many thousands or millions of millions of steps before the Lee Sedol match). If such a ridiculous behemoth could not beat Lee Sedol, Google would have been laughed off the planet.

Even then, as Bloemen hinted, AlphaGo behaved like a badly behaved beginner that wouldn't resign in game 4, despite the AlphaGo team containing a number of experienced players who could clearly see that the 1206 CPUs and the 4,937 MILLION searches per move were playing a string of lame duck moves and should resign. For me, this was the most revealing aspect of the entire match. One suspects that big money wanted to make a clean sweep to seal its advertising coup.

So, if 15 people + a zillion computer searches + a slicing and dicing of hundreds of thousands of the highest-level games from KGS/Tygem can beat one softly spoken and gentlemanly Korean guy, I don't think there is anything to celebrate here and certainly no 'breakthrough' in AI. If your calculator can multiply faster and more accurately than you can, so what? The real wonder is what a mindblowingly powerful player Lee Sedol is, to hold his own ground, game after game, in such an absurdly asymetrical contest. It's beyond belief. It's also a testament to the awesome majesty of Go that it requires such an arsenal of computing and manpower to get even an ill-mannered grip on the game.

To quote Emanuel Lasker, World Chess Champion for 27 years and recognized as one of the greatest Chess players of all time:
"While the baroque rules of Chess could only have been created by humans, the rules of Go are so elegant, organic, and rigorously logical that if intelligent life forms exist elsewhere in the universe, they almost certainly play Go."

What he means is that Go was not invented by human beings . . .


Really great post Stoner101,

Really great post Stoner101, but one minor correction: the quote is by Edward, not Emanuel, Lasker, who, according to Wikipedia, was nowhere near the strength and fame of Edward. Of course, it would be great to have go praised by a chess player considered to be the strongest ever. Smile

Thanks for the possible

Thanks for the possible correction, Decros.

I researched this today and was unable to confirm definitively one way or the other who said it, Edward or Emanuel. Wikipedia says Edward and gives the American Go Association as a source; the latter gives no source reference, so we have to take their word for it. If you or someone else has a source reference to something published by Edward with this quotation, I'd be happy to get it. I have my doubts.

What is certain is that Emanuel was also a strong Go appreciator and player. There is a nice photo of Emanuel playing against one of the top German Go players, Felix Dueball, in 1930, with what looks like a 'levitating' board held up by invisible aliens . . .
Closer examination shows the board is sitting on a largely hidden table, but the expression on the guys' faces is evidence that they are mesmerised. You can see at the same Senseis page that there is a discussion about whether it was Edward or Emanuel who had the trans-galactic thoughts about the game.

The same photo and the full record of a later (1931) Emanuel game against Dueball is on a German-language site devoted to Lasker:
Emanuel won by 1 point. Dueball was supposedly already the top player in Germany in the early 1920s and had spent a year studying in Japan, so Emanuel was obviously no slouch.

A number of German-language sites give the quotation as being from Emanuel and not Edward. Richard Bozulich's also gives (part of) the quotation on their home page as being from Emanuel, but there is no source reference. I would tend to trust Bozulich more than Wikipedia and the AGA, but that's just personal feeling. Unfortunately, I have been unable to find a copy of Emanuel's 1931 book, "Brettspiele der Völker" (Board Games of the Nations), which apparently has a substantial chapter on Go. That might be where the quotation is taken from? Perhaps a German reader of this forum can confirm one way or the other.

Given Emanuel's strong philosophical leanings and the fact that Edward's book on Go, "Go and Go-Moku" (1934, rev. ed. 1960), doesn't contain the quotation as far as I can find and is relatively sober about the game, I think it more likely it was Emanuel who said/wrote it. The German site devoted to him has a quote
saying that if he had encountered Go before Chess, it was unlikely he would have become World Chess Champion (too spaced out playing Go, I guess!).

Regardless of who said it, I think anyone who starts into playing 19x19 games very quickly gets the unmistakeable feeling that this game is very, very weird. It's a cosmos in itself, a higher intelligence that has very much to teach us . . .