Subcultural engagement for the 21st Century...
Barbelith is a new kind of community (find out more)...
You can login or register.

Underground > Laboratory >

Man vs. Machine: Kasparov wins round one.

09:48 / 27.01.03

Kasparov seems determined to prove that people can beat computers at chess and is playing a match against Deep Junior. It looks like he won the first round pretty convincingly. You can see the moves here.

Personally, I think that computers will win this contest eventually. But I think Kasparov's criticisms of Deep Blue were fair - they tweaked it mid-game to adjust to his style. However, if he loses to Deep Junior, it'll be obvious that machines really are superior. When it comes to chess, anyway.

10:10 / 27.01.03

Booyah! Take that, Skynet! The human race isn't redundant yet! Give us another decade or so...
Of course, the next project is to build a computer that can win a rap contest against Eminem. More 'now'.

12:17 / 27.01.03

It is, unfortunately, a fact that one day a computer that can beat anyone will be developed. Chess playing algorithms work on the principle of 'deep look-ahead' - that is every time the 'puter is required to make a move it calculates every single move to the end of the game and makes the one that is most advantageous to it.

Obviously, at any given time this number of moves is trillions, so computer chess players use shortcuts to lessen the amount of look ahead time. It is these shortcuts that (currently) allow human players to beat them. However, once chips are fast enough not to need these shortcuts, it is game over for the flesh.

I would guess one of the first (non military) uses for a quantum computer would be a chess machine...

13:27 / 27.01.03

Yeah, kinda. Except that brute force alone isn't enough. Or at least, not as you've described it. Computers also need to be able to trim the decision trees so as to choose a reasonable move and form some kind of prediction of the opponents next move(s). You need a fair amount of programming sophistication on top of your big processor.

Our Lady of The Two Towers

14:07 / 27.01.03

Is it in any way meaningful that a machine can crunch numbers better than several pounds of a substance that looks like porridge and originally evolved to try and work out how to eat without being eaten?

The Natural Way

14:55 / 27.01.03

Kasparov wins: BABALITY!!!

15:38 / 27.01.03

I am still loking forward to the day when Kasparov will be beating by a girl, preferably a 12 year old with clear signs of impending dykedom. I hate Kasparov and think he's a prick. There are many attitudes one can take toward chess, and winning by brute force is the least interesting of them.

Our Lady of The Two Towers

10:11 / 28.01.03

Are you saying Kasparov wins by brute force or Baby Blue?

The Natural Way

11:29 / 28.01.03

I don't know about fred, but I'm saying that his specialmove involves transforming the enemy into a baby.

21:56 / 28.01.03

And the second game ends in a draw. Hmmm. Looks like Kasparov can beat this computer.

Our Lady of The Two Towers

12:21 / 29.01.03

But who was it that forced the draw, Kasparov or the computer?

12:40 / 29.01.03

I've looked at the game a bit and Kasparov should have won - he made a mistake at a certain point in his attack which allowed the computer to force a draw. Given that he was playing black, I think it counts as a moral victory for Kasparov.

What I'm finding interesting are the reports that the current thinking on playing computers is to play them like humans. In the past, human players exploited weaknesses in the computer style - always going for a piece advantage, not dealing with sacrfices well, etc. But now it seems that computers are being afforded some respect in that players aren't looking for "tricks" with which to beat them. Which probably tells you that chess AI is about to supercede its human equivalent.

13:06 / 01.02.03

Here's some info on the draw and footage/reports for the next .

23:10 / 02.02.03

How does the fact that Kasparov made a mistake make that a moral victory for him?

23:21 / 02.02.03

Well, it was a moral victory in the sense that if he'd spent a few more minutes on that particular move he would have probably won. He essentially had the computer beat. He outplayed it...almost.

Of course, he lost the next game and the fourth looks like it is heading for a draw.

14:58 / 03.02.03

I'd call that a "near victory" rather than a "moral victory" meself. If he didn't win, he didn't win.

I've probably got a little behind on the current state of chess research, but I seem to recall that the human-esque strategies were being treated as a bit of a sideshow, more of use to psychologists trying to understand humans than in actually beating humans at chess. (The latter, of course, is a far more useless activity generally.) The point was that increasing hardware speeds amplified the effects of improvements in the algorithms. As you say, it's not just a question of crunching numbers, there's a lot more to it than that, the permutations are far too numerous - but the human-esque strategies don't gain nearly so much from the speed increases which are happening all the time. In effect the brute force strategy is being aided by a much broader research base.

Brute force strategies are also easier to research and formalise since they're more easily expressable in standard algorithmic terms. With your mimicry strategies, there's all the problems with trying to work out how humans do it and then convert that.

I foresee that brute force algorithms will be able to trounce humans before too long - but it may be the case that, when they start playing each other using similar hardware, it will be beneficial to incorporate mimicry-type elements, at least to an extent.

09:31 / 04.02.03

I may have given the wrong impression here, fridge. I don't think that Junior has any mimicry subroutines.

What I understand, from the commentary I've read, is that the decision tree pruning that most chess computers employ is somewhat transparent to GMs. Computers famously go for pieces rather than position, and lack objectives etc. I don't think that Junior is trying to play human, but may just be trying to play chess better.

The point is that Junior is good enough at chess so as not to seem so obviously a machine. Someone said that Junior passed the chess Turing test.

This makes sense to me. A good solution of the chess problem should be indistinguishable for a good GM without deliberate mimicry - though might surpass GMs, I suppose.

15:48 / 06.02.03

Anybody know what the latest is on the match? According to Chessbase.com's schedule, games 4 & 5 should have already taken place but I haven't seen any reports on it.

23:04 / 06.02.03

Its level pegging. Kasparov has won one, Junior has won one and the rest have been drawn. The chess base main page now has reports on all the matches.

To be honest, I think Kasparov is losing it.

pointless and uncalled for

13:15 / 08.02.03

Report from this mornings news is one win each and four draws that puts man equal to machine in the arena of chess.

I'm aware that we only field one champion and in theoretics there could be unlimited computer opponents but their difference is such that internally (as a subtext for the brain) they may as well be precisely the same machine.

On balance I see this as a far greater victory for machine, or far closer to reality the creators as it was demonstrable of two qualities.

Learning, which it performed at a level that exceeds human capacity and an absence of constitutional deterioration.

But for those of you concerned about of future that contains a rise of the robots as our masters, be aware that the maintenance of Deep Junior is far higher than that of Kasparov, who can make a cup of tea.

22:29 / 08.02.03

Wasn't Kasparov beaten by one of IBM's vacuum-tubed innard thugs before? Deep Blue wasn't it?