AlphaGo has won another 50 games against the world's top Go players, this time with little fanfare:
DeepMind's AlphaGo is back, and it's been secretly crushing the world's best Go players over the past couple of weeks. The new version of the AI has played 51 games online and won 50 of them, including a victory against Ke Jie, currently the world's best human Go player. Amusingly, the 51st game wasn't even a loss; it was drawn after the Internet connection dropped out. [...] Following its single game loss [in a match against Lee Sedol], DeepMind has been hard at work on a new and improved version of AlphaGo—and it appears the AI is back bigger, better, and more undefeated than ever. DeepMind's co-founder Demis Hassabis announced on Twitter yesterday that "the new version of AlphaGo" had been playing "some unofficial online games" on the Tygem and FoxGo servers under the names Magister (P) and Master (P). It played 51 games in total against some of the world's best players, including Ke Jie, Gu Li, and Lee Sedol—and didn't lose a single one.
That isn't to say that AlphaGo's unofficial games went unnoticed, though. Over the last week, a number of forum threads have popped up to discuss this mystery debutante who has been thrashing the world's best players. Given its unbeaten record and some very "non-human" moves, most onlookers were certain that Master and Magister were being played by an AI—they just weren't certain if it was AlphaGo, or perhaps another AI out of China or Japan. It is somewhat unclear, but it seems that DeepMind didn't warn the opponents that they were playing against AlphaGo. Perhaps they were told after their games had concluded, though. Ali Jabarin, a professional Go player, apparently bumped into Ke Jie after he'd been beaten by the AI: "He [was] a bit shocked... just repeating 'it's too strong.'"
Will there still be "Go celebrities" once DeepMind has finished mopping the floor with them and turned their attention elsewhere?
Related Stories
A year after AlphaGo beat the top Go player Lee Sedol, it is facing the world's current top player Ke Jie in a set of three matches (AlphaGo played five matches against Lee Sedol and won 4-1). AlphaGo has won the first match, so Ke Jie must win the next two matches in order to defeat AlphaGo. Although AlphaGo beat Ke Jie by only half a point in this match, edging out an opponent by a small margin is a legitimate strategy:
Ke Jie tried to use a strategy he's seen AlphaGo use online before, but that didn't work out for him in the end. Jie should've probably known that AlphaGo must have already played such moves against itself when training, which should also mean that it should know how to "defeat itself" in such scenarios.
A more successful strategy against AlphaGo may be one that AlphaGo hasn't seen before. However, considering Google has shown it millions of matches from top players, coming up with such "unseen moves" may be difficult, especially for a human player who can't watch millions of hours of video to train.
However, according to Hassabis, the AlphaGo AI also seems to have "liberated" Go players when thinking about Go strategies, by making them think that no move is impossible. This could lead to Go players trying out more innovative moves in the future, but it remains to be seen if Ke Jie will try that strategy in future matches against AlphaGo.
Although Google hasn't mentioned anything about this yet, it's likely that both AlphaGo's neural networks as well as the hardware doing all the computations have received significant upgrades from last year. Google recently introduced the Cloud TPU, its second-generation "Tensor Processing Unit," which should have not only have much faster inference performance, but now it comes with high training performance, too. As Google previously used the TPUs to power AlphaGo, it may have also used the next-gen versions to power AlphaGo in the match against Ke Jie.
Along with the Ke Jie vs. AlphaGo matches, there will also be a match between five human players and one AlphaGo instance, as well as a "Pair Go" in which two human players will face each other while assisted by two AlphaGo instances. This intended to demonstrate how Go could continue to exist even after Go-playing AI can routinely beat human players.
Also at NPR.
Previously:
Google DeepMind's AlphaGo Beats "Go" Champion Using Neural Networks
AlphaGo Cements Dominance Over Humanity, Wins Best-Out-of-5 Against Go Champion
AlphaGo Wins Game 5, Wins Challenge Match 4-1 vs. Lee Sedol
AlphaGo Continues to Crush Human Go Players
(Score: 3, Insightful) by tonyPick on Friday January 06 2017, @09:53AM
There are still "athlete running celebrities" despite the fact that they can be outrun by an inexpensive car.
The fact that AI algorithms and techniques have improved to "solve" a complex strategy game like Go is interesting, and the tools developed might give some insight into the game itself and associated tactics.
But fundamentally it's still interesting as a complex competitive contest between two people, and the best Go players are "celebrities" because they're better than other people, regardless of how well machines can do.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @10:23AM
A car is not a running bipedal robot, or running blade leg man, or running double artificial leg cyborg.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @01:55PM
That's true, and a car also can't run through terrains that a man can. Even among mammals, there seems to be the ability to go really fast for a moment like a cheetah or to go moderately fast for a long time like a man. I'm confident that Boston Dynamics will be successful in creating our robot cheetah-Tarahumara [wikipedia.org] overlords... just as soon as we get all those fantastic materials and applications graphene and carbon nanotubes are supposed to bring.
(Score: 2) by bob_super on Friday January 06 2017, @06:29PM
> Even among mammals, there seems to be the ability to go really fast for a moment like a cheetah or to go moderately fast for a long time like a man.
The wolves would like to show you how they hunt.
Just keep up with the bison, horses and wildebeests, and you might survive.
(Score: 1, Insightful) by Anonymous Coward on Friday January 06 2017, @11:07AM
People may become interested in the methods of solution rather than the solution. Much like mathematicians are more interested in the methods for solving x^n+2x+1=0 than the actual solution itself.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @05:33PM
Hey Ned, I'm gonna go solve some quintic equations.
(Score: 2) by Pslytely Psycho on Friday January 06 2017, @11:07AM
A better comparison would be chess.
Grandmasters have been getting beaten since 1996. There are still celebrity chess players.
Alex Jones lawyer inspires new TV series: CSI Moron Division.
(Score: 2) by opinionated_science on Friday January 06 2017, @01:38PM
An appropriate comparison. We know computers will beat humans as that is inevitable for any objective scored challenge.
I'm not so impressed by AlphaGo, unless it has rules that improve a normal (or even skilled) human.
The hardware however, is getting really interesting....
1PF for my desktop soon?
(Score: 2) by takyon on Friday January 06 2017, @09:18PM
It uses TPUs so FLOPS may not be the best metric.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: -1) by Anonymous Coward on Friday January 06 2017, @10:20AM
Can the AlphaGo get me free Bitcoin.
(Score: 2, Interesting) by Lester on Friday January 06 2017, @10:35AM
Aren't there chess player stars any more, Fisher, Karpov, Kaspárov etc.?
Well, no. Kaspárov was the last chess champion star. Nevertheless, there are still chess championships with a lot of followers. But it is a fact, great masters and champions aren't general public celebrities any more.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @10:41AM
Holy hell, really? What's next?? Classical music goes out of style??? We must do something about this! We need to get the kids drinking milk and listening to classical music and and playing chess again.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @10:43AM
Dope the milk with drugs and spent a night doing ultraviolence to old geezers! Let's see how many pawns fit up a pensioner's anus!!!
(Score: 1) by Lester on Friday January 06 2017, @11:46AM
I wasn't complaining, just telling that what happened to chess un going to happen to Go.
But if you change to this topic, I complain. It's sad that new generations are more interested in hype and stuff that requires low attention that in deep matters.
(Score: 5, Insightful) by Unixnut on Friday January 06 2017, @10:42AM
But it is a fact, great masters and champions aren't general public celebrities any more.
I feel that is more due to the fact that (in western society at least) anti-intellectualism is at the forefront. Being smart is just not something to aspire to. Just look at what passes for mainstream celebrities nowadays. Not one of them is a celebrity because of their intellect.
Most (from what I see) are celebrities because they are rich, or because of some scandal (sexual or otherwise), or because they are just good at marketing themselves.
From what I can see, the "quiet genius" is no longer celebrated at all, in any field really. They may get a note in the news from time to time if they do something globally impressive, but otherwise the majority of society goes "meh" and continues staring at the Kardashians.
(Score: 4, Insightful) by Anonymous Coward on Friday January 06 2017, @10:51AM
From what I can see, the "quiet genius" is no longer celebrated at all, in any field really. They may get a note in the news from time to time if they do something globally impressive,
"He was a loner, a quiet guy, never talked to anyone," said the neighbors as another domestic terrorist was dragged from his basement by the SWAT team. Is your quiet suburban neighborhood harboring terrorists who use Tor to spread their radical antigovernment views on metadata collection? Call the homeland tip line today or your credit score could be affected.
(Score: 2, Insightful) by Anonymous Coward on Friday January 06 2017, @11:13AM
Your "quiet genius" is now required to write grants that expire every 2 years and join committees to demonstrate sufficient service to the university. No-one is allowed to be quiet or have time to foment their genius. Get with the modern times, brah.
(Score: 1) by Lester on Sunday January 08 2017, @03:17PM
The chess lost its public interest after WW II. Alekhine was the last grandmaster. After him, chess turned a little into a team competition, with analysts etc. Alekhine and previous grandmasters, Capablanca etc, played alone with no more help than their brains.
After WW II chess was just another battle field un the cold war. Fisher was just pawn for USA government. And USSR dumped a lot of resources on chess (created chess schools, backed champions etc).
About the quite genius. It's not a problem of western culture, it's a problem of popular vs elitist culture. Old times culture was a few people's culture. I don't think a peasant in the middle age knew who Descartes, Goethe, Mozart was.
(Score: 2) by Bot on Friday January 06 2017, @04:54PM
Are there celebrities in the FPS videogame area? because none of them can beat the reflexes of a bot. It takes the very top players to beat a top level bot, usually because of some bug, sloppy programming. And those bots are not made to specifically win.
Account abandoned.
(Score: 2) by WalksOnDirt on Friday January 06 2017, @07:05PM
Have you never heard of Magnus Carlsen? He seems roughly as famous as Kasparov was.
(Score: 2) by HiThere on Friday January 06 2017, @08:23PM
In the US there were only chess stars for a few years, basically from when Fisher started his climb toward the championship until the guy who beat him lost it. And the peak was when Fisher was challenging for world champion. A year or two after then the "chess coffee-shop", HardCastle's, in Berkeley folded. It's been decades since I've seen a game played in public.
Partially this is due to increased competition. Computer games expanded during that period, and it also became a lot easier to play on-line...but that doesn't draw in kibitzers, so it doesn't add to the popularity. And partially it's because there's no longer nationalism pushing it. And partially it's because it's against the flow of the current decades, which have returned to their normal anti-intellectualism. (When I was in high school during the 1950's, the chess club was only of interest to a very few. No members were socially popular.)
If you look though US history, anti-intellectualism has been dominant though most of the time. Usually not strongly dominant, but still dominant. I think this last election is the first time it's been dominant enough to decide a national election, though.
Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
(Score: 2) by DannyB on Friday January 06 2017, @03:03PM
It is interesting to hear that AlphaGo is crushing human opponents.
Now that AI can solve this complex strategy game, how about getting AI to solve other complex strategy games like War. To a computer, war is just an abstract problem to solve. And "crushing" human opponents can take on a dual meaning of both (1) winning and (2) a means of winning.
The lower I set my standards the more accomplishments I have.
(Score: 3, Interesting) by mhajicek on Friday January 06 2017, @03:31PM
I seem to recall that someone is working on an ai to play Starcraft II with the same inputs that humans get. That's a big step in that direction.
The spacelike surfaces of time foliations can have a cusp at the surface of discontinuity. - P. Hajicek
(Score: 2) by takyon on Friday January 06 2017, @11:58PM
That "someone" is the same people who created AlphaGo.
DEEPMIND. DEEPMIND. DEEPMIND.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by looorg on Friday January 06 2017, @03:36PM
Overall I don't really find this all that interesting, such as how the novelty of Deep Blue once upon a time beat all the chess players. It was interesting when it happens but hardly a surprise that it (machine) would eventually do it. From what I understand it's essentially the same - a piece of software that evaluate ALL the moves, not even just the next move but many moves or counter-moves ahead up to the entire game, and then picks the best and it does this over and over and over again.
It would be like a mechanical punching machine hitting boxers in the face and smashing it in and then say that boxing as a sport is now dead due to machine being of superior strength and precision, or any other sport analogy - eventually we'll have robots that can run fast and better then humans all while and playing their game better to. I don't see anyone going to the robo-football-event of the week to see robots kicking (or tossing) the ball around.
Personally I found the tagline of the Arstechnica article more amusing and intriguing: Go stones will break my bones, but AI overlords will never harm me.
(Score: 2, Informative) by Scottingham on Friday January 06 2017, @04:02PM
You missed the point. Go is so complex that, unlike chess, all moves CANNOT be determined. That's why this is still big news even some months after the story broke.
I find it most interesting that less than two years ago we thought that it would take another 5-10 years before AI could beat a human at Go.
(Score: 2) by looorg on Friday January 06 2017, @04:40PM
I don't think I have ever read anything about GO having infinite moves. There is a limit, it's just a really large amount of them.
I do know that it's a more complex game then for example chess but there is still a finite amount of possible legal moves. A lot of those moves are not even going to be interesting moves, they might be legal but not interesting. As computers and software become better and faster that ought to diminish the complexity as far as the number of interesting moves goes. They might all have to be evaluated but that is just a matter of computation and that is all that this really does - it's massive amounts of computation far beyond the scope of a single human. Interesting in itself but not interesting enough for it to eliminate human Go-celebs.
Do you get excited to when you Sudoku solver does its brute-force or programmed attack? I give you that this, Go, is somewhat more sophisticated but in principle it's the same - a machine or a program excelling at the single task of its entire purpose.
(Score: 3, Insightful) by requerdanos on Friday January 06 2017, @06:06PM
I think you are talking about different things.
Even though Go by definition can't have infinite moves, that's doesn't mean that all the moves (and their value interpretations) are actually able to be determined by any known technology during the lifetime of a given human player/opponent--under which condition the moves aren't infinite, but still can't be determined.
The wise folks over at https://en.wikipedia.org/wiki/Computer_Go [wikipedia.org] give the following observations about AI trouble with just the endgame to illustrate some of the difficulty involved:
(Score: 0) by Anonymous Coward on Friday January 06 2017, @06:21PM
I don't think I have ever read anything about GO having infinite moves. There is a limit, it's just a really large amount of them.
There are arguably more possible configurations and moves than the number of atoms in the universe, so not infinite, but not something a computer can do either.
(Score: 0) by Anonymous Coward on Friday January 06 2017, @07:36PM
You don't bruteforce go, or chess for that matter. However, the thing with chess is that a human player can be beaten with an algorithm that gives same results as bruteforcing next n moves which is a pretty uncreative approach. In go, that's a no-go. You absolutely need to do something "smarter" unless you have a dyson sphere around the sun made of computronium.
(Score: 2) by takyon on Friday January 06 2017, @09:21PM
I might borrow that one.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 0) by Anonymous Coward on Friday January 06 2017, @10:26PM
unless you have a dyson sphere around the sun made of computronium.
I might borrow that one.
Want to mine the last bitcoin?
(Score: 2) by takyon on Friday January 06 2017, @10:33PM
I gotta hoard all the Gemcoins so I can become a billionaire.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2) by bzipitidoo on Friday January 06 2017, @05:09PM
We aren't gold medalists anymore in chess and now in Go. This has terribly upset us. The whole tone of this news is mournful and fearful. Soon AI will be better than us at every game. Noooo!!!
Get a grip. Games are in a way a huge simplification of reality, a better defined, rigid, and far, far smaller space very amenable to the sort of mechanized calculating that computers excel at. No one can match computers at mere addition and multiplication, computers can do millions of those per second, while a human would need lots more time just for the I/O. Just to read one multiplication problem and communicate the answer would take a person several seconds. No one is too troubled by computers' superiority at basic math. These games with very well defined rules are akin to basic math in many ways, and now with these powerful machines, we can explore and discover more about these games. For instance, we found that checkers is a draw with perfect play. Seems very likely the same is true of chess and Go, but until they are solved, we can't be sure.
I'm guessing that the way things eventually fall out will be that checkers, chess and Go are relegated to tic-tac-toe status. There's no point holding tic-tac-toe tournaments or a world championship, the game is solved, and simple enough that most people can just memorize the appropriate moves to make. The only wrinkle in tic-tac-toe is that the simple heuristic of always choosing to play in the center, and then the corners is wrong in one case, and will lose the game. Memorize the exception to that rule of thumb, and you too can be an unbeatable tic-tac-toe player.
Chess is a long ways towards that destination now. At grandmaster levels, chess devolves into an exercise in mental stamina and memorization. That explains why players are at their best in their 20s and 30s. Otherwise, how could a more experienced player actually do worse? How could a 50 year old player with 30+ years of experience not be as good as that same player 20 years before? It's like physical sports. Players peak when young.
That AlpahGo is finding "non-human" moves that are excellent shows that we've gotten stuck in a rut and are overlooking good strategies, haven't explored those corners of the Go universe. There was certainly some of that in chess. For instance, chess has this arbitrary 50 move rule to end games that otherwise might be endless, and thanks to computers, we know 50 moves is not quite enough. The king, knight and bishop vs king win can take a maximum of 57 moves.
This doesn't show that humans will soon be second rate at everything, that the Robot Apocalypse is coming. I'm guessing at a blended future in which our own intelligence is enhanced by computers implanted in our bodies, likely directly in our brains. We will become cyborgs.
(Score: 1, Informative) by Anonymous Coward on Friday January 06 2017, @07:29PM
It seems incredibly unlikely that perfect play would result in a draw in Go. White is given an extra half point advantage so there will not be a tie in points if the game actually ends.
However, this can still happen under the simple ko rule, which permits the board positions to have a cycle (thus the game never ends). Perfect play does not create this situation by accident, meaning that all possible alternate strategies attempted by either player (white or black) lead to inevitable defeat for that player.
So almost certainly there is a winner with perfect play in Go.
(Score: 0) by Anonymous Coward on Saturday January 07 2017, @03:26AM
I remember reading not that many years ago about how Go is so complex that computers would never be able to play it well enough to beat the best human players. Just too many possible moves to do an n-depth search, yadda, yadda. I'm sure they said the same about self-driving cars. And flying machines. And horseless carriages.
The world is full of "experts" who don't know shit...