World's top Go player loses second match of five against Google AI

[kurros]

Lee Sedol wins match 4 against AlphaGo.

http://www.ibtimes.com/lee-sedol-vs-alphago-after-three-straight-losses-korean-champion-strikes-back-maiden-2335472

Oh exciting! Not over for humans yet it seems.

[Lion IRC]

Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise.

AlphaGo is not doing that, which is exactly why it's news.

I thought it was news because Go has a phenomenally larger number of possible move combinations than chess.

But remember, it doesn't have to look at EVERY single possible combination - it only has to look at the winning combinations it had in its own data bank. And that data bank of probabilities was greatly enlarged because programmers deliberately developed a larger set of game outcome stats by having multiple computers playing one another over and over again at speeds much faster than a human could play.

Yes, the human who plays a huge number of games also gathers experience/memory but no human can match the brute force accuracy of memory and access speed in a timed game.

[Lion IRC]

Lee Sedol wins match 4 against AlphaGo.

http://www.ibtimes.com/lee-sedol-vs-alphago-after-three-straight-losses-korean-champion-strikes-back-maiden-2335472

Oh exciting! Not over for humans yet it seems.

Yes. Extraordinary that a human could even win/draw one game.

[Lion IRC]

I've worked for decades with various network typologies and search methodologies including bayesian networks and DAGs, they are all variations on the basic notions.

To the layman the statement "Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise."  is absolutely correct. It is also specifically correct in the case of AlphaGo due to the use of Monte Carlo.

Ok well I violently disagree with that and think it grossly misrepresents the situation, especially to laypeople.

I would love to see two computers violently disagreeing with each other.

[kurros]

Lee Sedol wins match 4 against AlphaGo.

http://www.ibtimes.com/lee-sedol-vs-alphago-after-three-straight-losses-korean-champion-strikes-back-maiden-2335472

Oh exciting! Not over for humans yet it seems.

Yes. Extraordinary that a human could even win/draw one game.

No, what was extraordinary was that a computer could beat a grandmaster Go player in the first place. AI researchers would be very disappointed to learn how trivial some members of the public think their achievement here is.

[Trinity]

AlphaGo would lose more often if it had emotions. Lee Sedol has said that the pressure made him nervous. Since Lee lost the game after 3 matches, the 4th match was less emotional and therefore easier to win.

[Lion IRC]

Right.
This cold, robotic indifference makes the program conspicuously LESS human.

[Lion IRC]

...AI researchers would be very disappointed to learn how trivial some members of the public think their achievement here is.

I did say;

"Neural network" is lovely jargon and I don't mind that it is used as a metaphor. I don't even mind a little bit of human vanity/ego getting involved and patting ourselves on the back at our own cleverness.

[wonderer]

Right.
This cold, robotic indifference makes the program conspicuously LESS human.

Of course AlphaGo is conspicuously nonhuman.  If you think anyone in this thread has suggested otherwise then that is just something else you have misunderstood.

However, the sort of information processing AlphaGo does IS much more like that which occurs in a human brain than is the case with earlier computers designed to play Go.  It is the human-like neural network processing which allowed AlphaGo to develop good intuitions about playing Go.

It is kind of amazing to me that you guys are so threatened by this, that you need to engage in so much denial.

[Trinity]

''It is kind of amazing to me that you guys are so threatened by this, that you need to engage in so much denial.''

I call it robophobia. The next step is AlphaGo demanding equal rights.

[aleph naught]

I've worked for decades with various network typologies and search methodologies including bayesian networks and DAGs, they are all variations on the basic notions.

To the layman the statement "Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise."  is absolutely correct. It is also specifically correct in the case of AlphaGo due to the use of Monte Carlo.

I agree with kurro, that's absolutely misleading to the lay-man. Even humans perform search trees when strategizing.

[kravarnik]

Lee Sedol wins match 4 against AlphaGo.

http://www.ibtimes.com/lee-sedol-vs-alphago-after-three-straight-losses-korean-champion-strikes-back-maiden-2335472

Oh exciting! Not over for humans yet it seems.

Yes. Extraordinary that a human could even win/draw one game.

No, what was extraordinary was that a computer could beat a grandmaster Go player in the first place. AI researchers would be very disappointed to learn how trivial some members of the public think their achievement here is.

No more surprising than the fact that space satelites can orbit the Earth, while human beings cannot. Or that an airplane can stay in the air longer than a human being can.

What's extraordinary is that particular human beings got the proper design in place to construct that. Not the fact that after it was designed to be so and so, it did so and so.

Or that industrial factories can produce more products than manual human labor.


So, that in of itself isn't too surprising. Perhaps, if we lived in an age where any high technology was missing. But, since we are surrounded by it, then this achievement doesn't seem something so extraordinary. I've seen other pieces of machinery that do certain things better than human beings do. Haven't you? I mean, from young age, for all my life - I've been surrounded by such fine pieces of machinery, so that's why I may be viewing this whole thing as... kinda ordinary.


As to the earlier point by wanderer and "AG intuition," let's not mistake a simulation with a genuine phenomenon, shall we? Just because we may simulate intuition(which we still don't know what exactly "an intuition" is, just how it practically manifests), doesn't make it genuine intuition. Anymore, than a novel has actual, genuine and living characters in it(who simulate life, just like we genuinely experience it).


If I wrote:

"kravarnik woke up at 10 A.M. and had a cup of coffee. Then he went to the bathroom. Then he went outside to hang out with a friend of his."

That simulation of some random day of mine, although seemingly very similar to how I ACTUALLY LIVE, doesn't mean the character of my story is an actual, genuine person, no more than the simulation of intuition in a robot pre-programmed to do so, means it's an actual intuition.

EDIT: Some helpful distinctions:

- artificial does not equal natural
- simulation does not equal genuinity genuineness

Simulating alternative reality doesn't mean that what takes place in the simulation is genuine at any rate(and that it is AN ACTUAL REALITY of some sort). Making an artificial limb does not mean it's THE SAME as a natural limb - perhaps, it has the SAME FUNCTION, but it isn't the same thing.

[RichardChad]

I've worked for decades with various network typologies and search methodologies including bayesian networks and DAGs, they are all variations on the basic notions.

To the layman the statement "Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise."  is absolutely correct. It is also specifically correct in the case of AlphaGo due to the use of Monte Carlo.

I agree with kurro, that's absolutely misleading to the lay-man. Even humans perform search trees when strategizing.

do you personally understand the difference between a search tree and a neural network? That would mean, can you articulate the fundamental difference without googling?

[aleph naught]

I've worked for decades with various network typologies and search methodologies including bayesian networks and DAGs, they are all variations on the basic notions.

To the layman the statement "Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise."  is absolutely correct. It is also specifically correct in the case of AlphaGo due to the use of Monte Carlo.

I agree with kurro, that's absolutely misleading to the lay-man. Even humans perform search trees when strategizing.

do you personally understand the difference between a search tree and a neural network? That would mean, can you articulate the fundamental difference without googling?

The fundamental difference?! They're nothing like each other. It's like you're asking me for the fundamental difference between a car an an apple. One is just a smart way of ordering objects so you can find what you're looking for in O(n log n) time. The other is a universal function approximator.

You haven't responded to what I said: even humans perform tree searches when strategizing. That the AlphaGo program uses them doesn't mean that's all it is. If it was, it wouldn't be able to win games against masters like it is doing. It is only able to win because it has a neural network that has learned how to play the game at the level of a master. In other words, a massively impressive feat of machine learning. It's learned things that most humans couldn't learn, or would take a very long time to learn if they could.

[RichardChad]

I've worked for decades with various network typologies and search methodologies including bayesian networks and DAGs, they are all variations on the basic notions.

To the layman the statement "Games are won by massive search trees. Computers will always be able to do this better. No news. No surprise."  is absolutely correct. It is also specifically correct in the case of AlphaGo due to the use of Monte Carlo.

I agree with kurro, that's absolutely misleading to the lay-man. Even humans perform search trees when strategizing.

do you personally understand the difference between a search tree and a neural network? That would mean, can you articulate the fundamental difference without googling?

The fundamental difference?! They're nothing like each other. It's like you're asking me for the fundamental difference between a car an an apple. One is just a smart way of ordering objects so you can find what you're looking for in O(n log n) time. The other is a universal function approximator.

obfuscation! man, you sure love em..

both have decision networks and both have transform functions (see red–black tree for example)

You haven't responded to what I said: even humans perform tree searches when strategizing. That the AlphaGo program uses them doesn't mean that's all it is. If it was, it wouldn't be able to win games against masters like it is doing. It is only able to win because it has a neural network that has learned how to play the game at the level of a master. In other words, a massively impressive feat of machine learning. It's learned things that most humans couldn't learn, or would take a very long time to learn if they could.

do you understand the difference between human learning and machine learning?