Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

AlphaGo beats Lee Sedol 3-0 [video]

youtube.com

Post reply on HN
Print view View original

301–310 of 428 posts

Re: AlphaGo beats Lee Sedol 3-0 [video]

#301

Post by aab0 »

bad_user wrote: post #151
alextgordon wrote: post #104

Earlier quoted context omitted.

What is our purpose if computers can do everything better than us? It feels like computers have taken one aspect of humanness: logic. Computers could do arithmetic, do algebra, play chess, and now they can play go. It hurts because logic is usually thought to be one of the highest of human characteristics. Yes computers might never be able to replicate emotion, but even dogs have that. There's still some aspects we h…

Reading this thread, I believe there's one aspect not discussed: in a battle between man and machine, it's debatable who wins and depends on the domain, but a man-machine combination always wins over both. On emotions, that's a characteristic of life. With the consciousness we possess, without emotions we would quickly realize that life isn't worth living. I doubt that a "true AI", one with consciousness, will want t…

"Reading this thread, I believe there's one aspect not discussed: in a battle between man and machine, it's debatable who wins and depends on the domain, but a man-machine combination always wins over both."

It doesn't 'always'. Advanced chess is already dead, and judging from the pro commentaries, they currently are worse than useless in an 'Advanced go' setting. That may change, but given how much faster computer Go is reaching superhuman levels than computer chess, the 'Advanced go' window may have already closed.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#302

Post by tel »

awwducks wrote: post #12

Perhaps the last big question was whether AlphaGo could play ko positions. AlphaGo played quite well in that ko fight and furthermore, even played away from the ko fight allowing Lee Sedol to play twice in the area. I definitely did not expect that. Major credit to Lee Sedol for toughing that out and playing as long as he did. It was dramatic to watch as he played a bunch of his moves with only 1 or 2 seconds left on…

I think from a technological perspective there's very little question that AlphaGo could play ko. I would have imagined that AlphaGo would be better at ko than most human players since it's a question of balancing risk across the entire board. Human players might be more likely to be exhausted and choose suboptimally by the calculation deciding between different stakes on the board, but MCTS will correctly optimize for the long term potential of each major branch in the game tree.

So I'd be very surprised if that turns out to be the trick. Things that are hard for human players are not at all necessarily AlphaGo's weaknesses.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#303

Post by jonknee »

threeseed wrote: post #181
Radim wrote: post #110

Earlier quoted context omitted.

Alas, time is not on meat's side in this one.

It could one day be possible. There are savants that have the ability to work with complex working sets e.g. reciting Pi to a ridiculous precision, multiplying very large numbers or recalling images with photographic detail. So perhaps the problem isn't our brain's hardware as such but the operating system that runs on top of it.

> There are savants that have the ability to work with complex working sets e.g. reciting Pi to a ridiculous precision, multiplying very large numbers or recalling images with photographic detail.

Yes, but all computers do those things millions of times faster than even savants. And the computers are getting faster at it every year, savants today aren't any more clever than savants 100 years ago.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#304

Post by YeGoblynQueenne »

Eliezer wrote: post #79

My (long) commentary here: https://www.facebook.com/yudkowsky/posts/10154018209759228 Sample: At this point it seems likely that Sedol is actually far outclassed by a superhuman player. The suspicion is that since AlphaGo plays purely for probability of long-term victory rather than playing for points, the fight against Sedol generates boards that can falsely appear to a human to be balanced even as Sedol's probabili…

>> AlphaGo's core is built around a similar machine learning technology to Deepmind's Atari-playing system - the single, untweaked program that was able to learn superhuman play on dozens of different Atari games just by looking at the pixels, without specialization for each particular game.

Woa there. The Deepmind Atari-playing AI was too specialised for each particular game. It had a reward function that translated the score for it. It couldn't learn the importance of the game score on its own, just from "looking at the pixels" and it couldn't learn the significance of the score display and how it changed as a result of its actions on its own. All this had to be hand-coded. And if I've seen a hint that this means it couldn't really-really generalise to other games, like Deepmind claimed, then that's this bit that you report yourself:

>> Deepmind ... did reuse the widely known core insight of Monte Carlo Tree Search.

That is no mere detail. That is the crux of the matter, right there. Deepmind used their architecture to improve MCTS far enough that it could beat Lee Se-Dol. They didn't just add MCTS to their already general-game playing system.

Because the original Atari-playing AI was completely useless for playing Go, a game that doesn't have a score and looks nothing like an Atari game. So it wasn't very general at all, despite Google's and DeepMind's claims to the contrary.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#305

Post by jonknee »

YeGoblynQueenne wrote: post #205
quanticle wrote: post #166

Earlier quoted context omitted.

I find it funny that whenever there's a case of computers being able to do something that they couldn't before, whether it's drive or beat humans at Go, the goalpost on what is "true AI" shifts to be something that computers can't do yet. So let me ask you this. What would you consider to be "true AI"? At what point are you willing to say, "Okay, that's it, computers are just plain smarter than we are?" Because, fran…

>> What would you consider to be "true AI"? At what point are you willing to say, "Okay, that's it, computers are just plain smarter than we are?" Look, there's no doubt that computers can outperform humans in specific tasks. There's no doubt that AlphaGo is intelligent when it comes to Go, but on the other hand it would be completely incapable of tackling a different congitive task- say, language, or vision, or disc…

Humans have been trained for all the different scenarios you cite, AlphaGo has just been trained for go. Give AlphaGo 20 years to chew through training data and I think it would destroy a college sophomore at cognition.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#306

Post by tel »

Yuioup wrote: post #53
starshadowx2 wrote: post #36

Earlier quoted context omitted.

In game 2 there was a point where Michael Redmond seemed to do a triple take and couldn't believe the move AlphaGo played.

Yeah they seem to forget that Alpha-Go is looking deep into the future. I have not read the Nature paper but I assume it's playing out possible moves way into the future. At some point it figured that the Ko fight at the bottom was already won. Hence that white move at the top which nobody saw coming. Another interesting moment was when Michael Redmond said "A human would typically not spend too much time thinking on…

It plays out thousands of moves all the way until the end of the game using its neural net to quickly sketch optimal play. That's the big advantage! Humans read moves until they feel an outcome is favorable. AlphaGo reads out till endgame and plays moves that optimize for a win.

Also I'm not entirely sure how AlphaGo's time management works, but it's doing the same thing for every move—populating the game tree as deeply and intelligently as it can. It may just look for "30 seconds" on every move and then take the best bet meaning it's a more thorough and exhaustive reader than any human.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#307

Post by Shorel »

alextgordon wrote: post #104
pushrax wrote: post #31

It's important to remember that this is an accomplishment of humanity, not a defeat. By constructing this AI, we are simply creating another tool for advancing our state of being. (or something like that)

What is our purpose if computers can do everything better than us? It feels like computers have taken one aspect of humanness: logic. Computers could do arithmetic, do algebra, play chess, and now they can play go. It hurts because logic is usually thought to be one of the highest of human characteristics. Yes computers might never be able to replicate emotion, but even dogs have that. There's still some aspects we h…

Future research will be along these lines.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#308

Post by tel »

badminton1 wrote: post #84
flyingbutter wrote: post #50

The Chinese 9 Dan player Ke Jie basically said the game is lost after around 40 mins or so. He still thinks that he has a 60% chance of winning against AlphaGo (down from 100% on day one). But I doubt Google will bother to go to China and challenge him.

AlphaGo has also improved very quickly. Without doubt, the AlphaGo seen playing against Fan Hui would have lost against Lee Sedol. But in a couple of months its playing level raised significantly. Lee Sedol said he could beat AlphaGo, based on the Fan Hui games. Ke Jie said he could beat AlphaGo, based on the Lee Sedol games. Ke Jie belongs to a similar category than Lee Sedol, and we could see how Lee Sedol was comp…

While AlphaGo has been improving, it's also a little hard to see whether AG playing Fan Hui wasn't already strong enough to defeat LSD. MCTS picks the move most probable to win and if AG gets an early advantage it'll play slack for the rest of the game taking moves which never reduce its win rate even if they look boring and weak.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#309

Post by tel »

schoen wrote: post #59
Yuioup wrote: post #53

Earlier quoted context omitted.

Yeah they seem to forget that Alpha-Go is looking deep into the future. I have not read the Nature paper but I assume it's playing out possible moves way into the future. At some point it figured that the Ko fight at the bottom was already won. Hence that white move at the top which nobody saw coming. Another interesting moment was when Michael Redmond said "A human would typically not spend too much time thinking on…

> I have not read the Nature paper but I assume it's playing out all possible moves. To some relatively small depth, right? I hear the estimate that all possible moves in a Go game probably can't be physically represented in the universe (unless we learn much more about the structure of games' evolution).

No, it plays deep but only so broad. It uses a neural net (which playing by itself without MCTS already beats Pachi with like 80% probability) to sketch out the best moves until the end and then rates each move on its chance of winning.

This objective function is why Go playing AI jumped hugely in the last 10 years.

Re: AlphaGo beats Lee Sedol 3-0 [video]

#310

Post by moyix »

YeGoblynQueenne wrote: post #304
Eliezer wrote: post #79

My (long) commentary here: https://www.facebook.com/yudkowsky/posts/10154018209759228 Sample: At this point it seems likely that Sedol is actually far outclassed by a superhuman player. The suspicion is that since AlphaGo plays purely for probability of long-term victory rather than playing for points, the fight against Sedol generates boards that can falsely appear to a human to be balanced even as Sedol's probabili…

>> AlphaGo's core is built around a similar machine learning technology to Deepmind's Atari-playing system - the single, untweaked program that was able to learn superhuman play on dozens of different Atari games just by looking at the pixels, without specialization for each particular game. Woa there. The Deepmind Atari-playing AI was too specialised for each particular game. It had a reward function that translated…

I think you're wrong here. From the Nature paper describing AlphaGo:

"We also tested against the strongest open-source Go program, Pachi, a sophisticated Monte Carlo search program, ranked at 2 amateur dan on KGS, that executes 100,000 simulations per move. Using no search at all, the RL policy network won 85% of games against Pachi."

AlphaGo does use MCTS, but it seems that most of its improvements are actually coming from the deep reinforcement learning approach.

Post reply on HN