Are machines conceded to be more of less unbeatable at this point… or not?
Surprisingly to me, the question seems to still be open (from Wikipedia article on Deep Blue):
It also seems there have been no other grandmaster-vs.-supercomputer matches since Deep Blue’s 1997 win over Kasparov. If that’s correct, ISTM that the supercomputers haven’t yet built up enough of a track record against human grandmasters to claim dominance.
Although … it does appear clear that an average computer can consitently beat an average person at chess.
ALL HAIL OUR SOULLESS CYBERNETIC OVERLORDS!
http://en.wikipedia.org/wiki/Man_vs_Machine_World_Team_Championship
Well, not all Hail. But 8 to 4 Hail.
I’m pretty sure that’s false. I recall reading several news stories regarding the performance of various programs against grandmasters, though I don’t recall the details.
I stand corrected. I assumed there’d be links to subsequent developments – or at least mentions – on the Deep Blue Wiki article. Wiki’s not bad, but not perfect.
Damn … last year, the humans could do no better than draw, if I’m looking at those results correctly.
The machine’s programmers had the right to adjust the program between games, as Kasparov knew before the match. In no way was this unfair. The human player can always adjust it’s play to fit the machine (which Kasparov did), so why is it unfair for the machine’s programmers to do likewise?
Given Kasparov’s previous behavior, I silently predicted to myself that if he was defeated by a machine, he would insist that he had been cheated. And that’s exactly what happened. He has a history of making accusations that are simply untrue.
It was inevitable that machines were to become better at chess to the point that they could beat even the world’s best players. I was surprised that Kasparov would be the first world champion to lose to a machine, given his incredible skill and determination. But since these things were getting better and better, it was really a question of time.
It’s important to note that computers are winning not because of some incredible intelligence that we have managed to program into them, but because they are programmed with specific strategies that have been developed by humans combined with brute strength computational power. Brute force computation is the one thing a computer does really well. Computers are tricky for a grandmaster to play, because they don’t play like a human. Computers have about as much intelligence and strategy as your average stick of butter. Computers aren’t winning chess games because they are out-thinking humans, computers are winning because they are using raw number crunching ability to overcome human thinking.
Computers may be winning, but their chess playing ability still sucks.
It’s not a question of fairness – the human tinkering just makes me think that Kasparov actually lost to humans, not a machine. Now … if the machine were able to modify its own program between matches on its own, that would be a completely different matter.
I can’t see how you’re coming to this conclusion. Unless you’re referring to aesthetics, i.e. how elegant or beautiful the style of play is, then clearly the computers’ chess playing ablility is superlative. It’s like watching a human sprinter race a Ferrari. “Sure, the Ferrari left the human in a cloud of dust, but it’s still a terrible runner…”
Was Kasparov allowed to consult others about the machine’s play? If so, then it sounds fair. If not, not fair, the machine should have to play on its own merits as well.
The other point I’ve heard mentioned before is that the computer got to study Kasparov’s previous games, whereas Kasparov had no idea what to expect from Deep Blue. That seems like a pretty big advantage.
engineer_comp_geek is right in the sense that computers playing chess are just playing out all combinations of moves for the next X number of moves and choosing the move with the most chance of resulting in a win. It’s mostly brute force.
Some of the more interesting methods of solving problems, including game play, that don’t rely on brute force are genetic programming, genetic algorithms and neural networks. Someone did a checkers program developed through genetic programming that works pretty well.
Seems to me bottom line is humans still have the edge on the machines…sort of.
The very best human players will tend to beat the very best chess programs.
Most people (read average) will be beaten by off the shelf programs handily.
Thing with Kasparaov and Deep Blue is you had a hugely expensive and powerful computer complete with a team of programmers and chess experts to twiddle with it and design it to specifically beat Kasparov. While this is all well and good it hardly points to the machines having surpassed humans in chess. Just one machine goosed to the max and handheld the whole way.
I really wonder if Anatoly Karpov (or pick your grandmaster) sat down and played the same program Kasparov did if Deep Blue could have beat him.
In short we have yet to see a commonplace program/computer that can consistently beat any human. No doubt one day that will be the case but it is a ways out there yet I think.
This argument comes up often, and it depends I think on what people mean by chess-playing ability. The big difference between computers and humans is that computers never blunder in the chess sense; they always play at their full strength.
Kasparov played a 4-game match against the program Junior a few years ago. They each won once, and they drew the other 2 games. In the game Kasparov won, he just outplayed the computer for the whole game. In the game Junior won, Kasparov was outplaying the computer but made a quite simple blunder and lost. So by this result they are of equal strength, but looking at the games many would say that Kasparov is stronger.
Also, in correspondence chess, where blunders are basically eliminated, computers do quite poorly as well. The chess machine Hydra recently won a match against Michael Adams, who is in the world’s top ten, by a score of 5.5 - 0.5. However in correspondence, Hydra has lost several games.