10 Years After Big Blue Beat Garry Kasparov 368
Jamie found another MIT Technology review story, this time about Chess, Supercomputing, Garry Kasparov, and trying to make sense of just what exactly it all meant when a computer finally beat a grand master. An interesting piece that touches on what it means to play chess, the difference between humanity and machinery and how super computers don't care when they are losing. Worth your time.
What is "intelligence" (Score:5, Insightful)
I did an AI degree in the mid 90s and one of the things we covered was the definition of intelligence. After running through a few unsatisfactory definitions, my conclusion was that people used intelligence to mean whatever could be done better by a human being than anything else...
Actually, my favourite definition of intelligence, partly because of its succinctness, is "productive laziness".
Peter
It is a game of logic (Score:4, Insightful)
Oh, please. The hubris is overwhelming.
I play the game. I am not a great players, but it is a fun diversion and can help to develop focus and thinking skills. But, please, to say that Chess could have been beyond a computer? That is small, ignorant thinking.
The human brain excels at pattern matching in massive parallelism. It is this advantage we have over our current computers. But, new computer designs have gotten fast and with lotsa memory and storage space. It was only a matter of time until a computer had the right amounts of that speed, memory and storage space, coupled with programmers to make the best use of it and then no human would ever stand a chance.
As we get better with fuzzy AI type stuff, even games like Poker, Texas Hold 'em and others will even fall from our human hands.
The intuition we exercise is some random choice being made, but based on experience and a factor of acceptable risk of failure.
Some of the problems. (Score:3, Insightful)
The Computer doesn't care it is just focusing on the game 100% it is not even conserned if it is breathing or not overheating or a person behind it with a gun to shoot it if it looses. It is just running a set of processes, and using its memory to play the game.
Re:What is "intelligence" (Score:5, Insightful)
True, but I think that's just a special case of the general rule that, "People don't like when their expertise is systematized so that others can easily gain it." (Probably a better way to say that.)
This article would be more relevant if (Score:5, Insightful)
After it was discovered that IBM was tinkering using chess experts (that is, humans) to tinker with its software between matches, they're personae non gratae in the chess world now.
Chess is a bad example of thinking (Score:1, Insightful)
At each step in a game of chess, there is a finite, pre defined set of legal moves. You have up to 16 pieces you can move, and depending on their position, each piece has a number of pre-defined legal moves. As you try to look forward, the number of possible moves increases exponentially, but no matter how far forward you look, there is still a finite number of pre-defined legal moves.
No room for creativity at all. And that's where our difference comes in.
Computers are excellent at searching through a finite space of pre-defined values, which we in general suck at. On the other hand, we are excellent at coming up with creative solutions, where as the computer sucks at this (or rather, is completely unable to). That's why once the computer becomes fast enough, the computer will always win. Always. We are not there yet, but the computer is already winning most of the time.
What we are seeing in a match between Deep Blue and Kasparov is NOT a computer doing what humans are good at. It's a human doing what computers are good at. Kasparov has played for years to get as good as searching through a finite space of pre-defined values, where as computers have been doing this since day one. For the computer, the only difference is speed. For us, taking a mediocre player, and making him a million times faster is not going to make him play better. He'll just get beaten a million times faster.
Playing games like Chess (or even Go) is not the way to prove that we are more intelligent than computers. However, either game can be used to do exactly this. How? Not by playing...
Who came up with Chess?
When is the last time the computer came up with a game on it's own?
Re:A chess player's take on this (Score:3, Insightful)
Re:A chess player's take on this (Score:2, Insightful)
Re:Some of the problems. (Score:4, Insightful)
and more to the point, the computer doesn't even know what chess is. It's just adding, subtracting, fetching instructions from memory, etc. It's kind of like how a guy in a box doesn't really understand chinese, or how none of your brain cells actually know what slashdot is.
I wonder if it would be more accurate to say that a system, which included a computer as one of its parts, but also included a human programmer, beat Kasparov. Kind of like how it's not accurate to say that a few neurons and muscle fibers posted to slasdot. My brain cells and my fingers don't know what they're doing, any more than Big Blue knew what it was doing.
Re:It Didn't Mean Anything... (Score:5, Insightful)
1. It's almost trivial to make a program 'learn' from mistakes. Just store some negative value for that specific decision-point. Depends on your definition of 'learning', of course. But the principle is the same in humans and AI
2. Kasparov also adjusted his style (i believe there are certain playing-styles that are beneficial when playing against an AI), and i bet he had coaches and consultants
3. So what?
4. See above.
My point is that every time some AI people actually manage to out-do humans, humans tend to re-define what intelligence is. I bet if you'd tell somebody 100 years ago that a machine would be the world's best chess player, that alone would have been enough to consider the machine 'intelligent of sorts'. But as soon as we know how it works, it somehow looses the right to be called 'intelligent' (mechanical turk). I think this is because it seems to hurt humans that AI shows them that whatever gives us the right to call ourselves 'intelligent' is nothing more than the result of zillions of relatively simple interactions of little protein-machines.
IIRC (its been a while) the best way to determine what language a given text is written in, is amazingly 'stupid': just compare the ratio of how many times the different characters appear. The result is still amazing and should be considered 'kind of intelligent'.
So, just give AI some kudos, accept that there's a lot left to be done, and that the heuristics dint really matter, as long as the result is cool. (and please dont give me none of that Chinese Room Argument crap)
Re:What is "intelligence" (Score:5, Insightful)
For instance, how do you see a trail as it winds over grassland and leads into the woods? How does one see a year old trail that is partially overgrown, or a new trail not completely tramped down. How do you track down an animal from smattering of scat, nibbles and tracks over rocks, dirt, grassland, and the tree line? How does a human being see a camouflaged predator slinking behind the tree line? How do you read the sky and know what the weather will be later that day? How do you look at a river and know if it's crossable or not? Back at home, how do you play your relatives, friends, and enemies in the tribe so that you are elected leader when the Big Man passes away? Or how do you manage to convince your husband that your new pregnancy is his, and not your secret lovers'?
Computers seem to be like idiot savants. They are good at logic puzzles, things like factoring large number or memorizing the phone books. That's a very useful tool in our technological society, but I don't think it's the basis of human intelligence. Like some Autistic person, computers suck as the basics of social interaction, which any three year old understands the basic concepts of. I remember my friend's three year old putting on her parents clothes and getting dressed up when she heard that her parents were going to a Halloween party -- all without prompting. What kind of intelligence do you need to understand the concepts of 'a party' or 'dressing up'? Or simple thinks like standing on two legs or filling a glass of water -- never mind hunting and eating another animal, or following a trail.
Re:What is "intelligence" (Score:2, Insightful)
The difference between human and machine intelligence currently seems to be that humans use a variety of techniques, while computers tend to use a more limited set of techniques that are carefully tuned and tested. Human intelligence generally comes from using and cross-feeding different techniques in a complimentary way. The various kinds of intelligences include:
* Physical modeling
* Analogy application
* Formal logic
* Pattern recognition
* Language parsing
* Temporal Simulation
* Memory
* Others that I forgot
A.I. has not found a good way to combine bunches of these techniques together to feed back on each other, test each other, and triangulate an answer. Our definitions of intelligence seem to be shifting toward this "synergy" view.
Re:the supercomputers advantage... (Score:1, Insightful)
every chess novice learns that you play the board, not the opponent
the instant you make a potentially inferior move to sucker your opponent, you deserve what you get
Re:What is "intelligence" (Score:3, Insightful)
I am a Bolo Mark V of the line. Bolo uses radar, lasers, sonar, GPS, and satellited imagery.
How does one see a year old trail that is partially overgrown, or a new trail not completely tramped down.
Bolo does not care. Bolo tank treads crush all terrain obstacles. That which cannot be overcome is destroyed with main gun.
How do you track down an animal from smattering of scat, nibbles and tracks over rocks, dirt, grassland, and the tree line?
Bolo uses satellites to track live infrared images of animal's body heat. If animal give off scent, my chemical warfare receptors will identity as such. If animals hides under cloud cover or bush I will use aerial drones to seek it out. If I cannot find it, I will use napalm to reduce the cover first.
How does a human being see a camouflaged predator slinking behind the tree line?
Bolo does not have natural enemies.
How do you read the sky and know what the weather will be later that day?
Bolo is unaffected by weather conditions. However, Bolo is aware of weather by satellite imagery and understands who to use weather formations for tactical advantages.
How do you look at a river and know if it's crossable or not?
Bolo is completely submersible.
Back at home, how do you play your relatives, friends, and enemies in the tribe so that you are elected leader when the Big Man passes away?
Bolo only takes orders from supreme command. But Bolo unit is completely independent and can operate on remote planets without guidance for centuries.
Or how do you manage to convince your husband that your new pregnancy is his, and not your secret lovers'?
Bolo does not produce offspring, but if it did it would use DNA technology. Then it would destroy the opposition with main gun.
(I know I'm responding your questions with a famous fictional AI, but I think that is how it would deal with the questions)
Re:What is "intelligence" (Score:3, Insightful)
Re:the supercomputers advantage... (Score:4, Insightful)
Why is this "highly suspect"? I suppose you might think so if you made the mistake of believing that Kasparov was actually playing against a piece of hardware (the "computer"); but of course he wasnt. Kasparov was playing against a team of chess-knowledgeable programmers; Kasparov was playing against software. The only remarkable thing about the computer itself was its speed--it was fast enough to carry out the laborious recursive brute-force searches for optimal moves in about the same time as a human player would take to decide his move. In theory, you could have done the same thing with a 70s era computer...but the game would have taken forever.
I'm not a chess player, but it's my understanding that during important tournaments, players often talk to advisers to determine their strategy in the next game against a tough opponent. How is this different from the programmers tweaking the software between games? Fundamentally, this was a contest between Kasparov and a team of programmers. Kasparov surely knew that, and accepted the match under those conditions. So I don't think the IBM team can be accused of "cheating".
The fact that such accusations have been made shows how people--even the paranormal crowd that posts to /.--easily forget how computers and computer software work. Once you remind yourself that this is not a case of "man vs. machine", then the philosophical significance of the contest wanes. Computers do not play chess...only people do.
Re:the supercomputers advantage... (Score:2, Insightful)
Re:A chess player's take on this (Score:1, Insightful)
It's a little like saying: There are more natural numbers than their are atoms in the known universe, so there's no way [anything computational] could ever prove that there is no highest prime number. Turns out, that there are mechanisms far more effective at this sort of thing than brute force.
Similarly for chess algoriths. Whole swaths of those 10^105 operational positions can be dismissed based on the fact that they would never maximize the probability of winning (e.g. immediately pulling out your queen and putting her in threat from the opponents pawns...bad idea, making all future positions irrelevant)
Re:What is "intelligence" (Score:3, Insightful)
Really, Turing was saying that if you can't distinguish between entities you consider intelligent and those you do not solely by communicating with them, your definition of intelligence is flawed.
Doing that communication in text in English is not a requirement, unless you want to actually perform the experiment, which was not the point for Turing. For what it's worth, various people have done so, generally giving the potential AIs some sort of (enormous) leg up by restricting the subject matter of the conversation, and the AIs blow it anyway; they're nowhere close.
A binary test for intelligence will prove illusive as intelligence is not a binary quality, nor even a well defined one.
While some people might actually claim jaguars were not intelligent in the way humans are, the Turing test does not. Jaguars cannot communicate in any way that would allow the test to be performed in the first place. Assuming we had some meaningful way of communicating with jaguars, I would guess that I could tell whether I was communicating with a jaguars or a human. If that were the case, then the Turing test would not reject a definition of intelligence that distinguished between jaguars and humans. That's all the Turing test does, is reject definitions of intelligence that distinguish between entities that you can't distinguish between via communication.
As for hunting, I reject any test of intelligence that says some ants are intelligent but Steven Hawking is not.
Let me explain this as a chess player... (Score:2, Insightful)
No chess program that is only programmed with the moves and rules could ever beat a Grandmaster, International Master, or even a rated Master here in the USA. Ever. A chess program with the moves, rules AND Alpha-Beta search (that counts only the value of the pieces) might beat a Master occasionally on a very fast computer. Let me hash this out in the three stages of a chess game:
OPENING: Kasparov was beaten by another human who programmed 200 years of already-known opening moves into Deep Blue. Indeed, the last game was lost by Kasparov because he fell into a "known" opening trap. "Known" to the computer through human programming.
MIDDLE GAME: Strategy used by DB was programmed by a human Grandmaster. Then it is checked during the game by brute-force calculation. DB played roughly equal here. Every Grandmaster today checks his middle game ideas for tactics or traps with a computer that can find mistakes in a few minutes that would take a human days or weeks to find.
END GAME: Knowledge of end game positions and strategy was programmed by a human Grandmaster. Kasparov outplayed DB completely here.
No machine or software has ever beaten a Grandmaster without human intervention.