Why did IBM retire it?

I would like to see world champion or world champion contenders challlenge it every year in World Championship match conditions.

Computers CHEAT.

They have tables they can use to LOOK UP OPENINGS. They also LOOK UP ENDGAMES in tables.

Computers should not be allowed tables at all except a modest opening system.

I'm not so sure that a human could beat a computer, even at postal speeds any more.  During that amount of time every possible chess game from a given point could be calculated by the computer.  Draws would be common but an actual win against a computer specially designed for chess would be next to impossible to beat.

And when it comes to computers driving cars, the biggest problem they have is reacting to humans that are irrational, easily distracted and quite unpredictable at times.

I believe it  is likely that the World's Champion Postal Player would beat any computer at postal time limits. Computers have an advantage at regular over the board speeds with calculations of millions of positions instantly and access to an enormous opening book and database. It is an unfair contest under OTB tournament conditions. The computer advantage would be negated at postal speeds. That is a contest I would welcome!

The problen with a computer driving a car is not the computer its the sensors required to give the computer information about its surroundings cannot compete with our senses and sensory/brain connection. A computer cannot tell the differance between a boy chasing a ball and a car jacker and would probably run you over thinking you were a chess player.....

For all the doomsayers I have just two things:

1) Humans may have a huge calculation possibility, but most of that is used on simply stying alive and walking upright.

2) Robots can't do those things yet. Let alone do anything remotely like thinking.

If there ever will be terminators and such it's gonna be when we're all long dead.

According to wikipedia it was a joke.

On the TV show QI, Stephen Fry joked that part of Deep Blue was currently being used as a booking clerk for United Airlines.

Computers are tools period. They will get stronger and faster but they will always be tools for real chess players to use and make themselves better at chess.

I'd say Kasparov won a few games not because art is unpredictable, but because chess strategy is difficult to program well. 

The brute force algorithms usually used by chess computers makes them tactical geniuses - they can't miss a tactical combination many, many moves away.

But strategically, the program usually just follows a series of additional weights based on strategic considerations.  Honestly, I don't think that Deep Blue had much of a strategy at all - very tactical machines will, in a way, just move their pieces around until the opponent makes even a slight tactical error, then capitalize on that.

While some chess programs do have large amounts of code dedicated to strategy (check out Crafty's code for pawn structure, among other things), I feel that the strategic chess machines in the near future will be chess program using neural networks or other machine learning schemes.

I feel a neural network could be trained to be 'close enough' to a grandmaster to play excellent strategic chess, based on the situation, but without an incredible amount of training, it might allow tactical weaknesses to unexpected moves.

The solution, of course, is to somehow weight or integrate machine learning with tactical brute strength - to merge the current strength of chess software with the human aspect it's been missing.

42

According to QI, a British television programme which is based on "Quite Interesting" things, and myth-busting, Deeper Blue (the first computer to beat a chess grandmaster) is now working for United Airlines as a reservations clerk.

This is an interesting topic. Computers are evolving rapidly as we all know and the debate around AI and supercomputers "intelligence" is puzzling the mind of scientist, poets and well......chess players.

The way we play chess today is quite different from the past thanks to computer analisys of openings and endgames. So this machine can "think" but in a way wich is different to the human brain. They use brutal force and calculaton strenght wich are definitely helpfull for us. They can even pass the Turing test, (for the average chess player it's almost impossible to say if your opponent is a human or a computer nowdays..... so be advised chess player at chess.com ;-)

So why Kasparov still wins over machine? Because chess is math and geometry among many other things but is also an art...... and art is unpredictable.

So Deep Ice isn't really anything like Deep Blue.

The main computer part of Deep Blue could probably be beaten by a modern PC.  The really powerful part of Deep Blue which allowed it to beat Kasparov was the chess hardware specifically designed to quickly calculate chess positions.

Similar to how modern gaming computers have graphics cards to quickly compute calculations for 3D graphics, Deep Blue had a powerful array of hardware chips designed for the sole purpose of calculating a chessboard in parallel, giving it a huge speed boost (and therefore, calculating power) over even the strongest chess programs that run on a single processor.

Unless the silicon chessboard used by Deep Blue is somehow reusable, it's unlikely that Deep Ice functionally has anything in common with Deep Blue.

Separately, responding to the comment about computers driving cars...that has happened to.  The most recent project to autonomously drive a car that I've heard of uses artificial neural networks - an electronic design that forms a very crude model of synapses and human brain connections.  Given a set of training data, ANN's can "learn" behavior by adjusting a very large number of variables (weights on the connections between neurons), breaking a complex nonlinear problem into manageable functions.

I wouldn't trust a computer driving my car today, but it's certainly feasible that a good neural network design with ample training data would allow a computer to drive a car at least as well as some other drivers on the road.

computers can play chess well but they can't drive a car good at all even with a supercomputer. Probably with the right software they will but in real world situations they can really make bad mistakes even so

The human brain makes 10 times as many calculations per second as blue ice and it takes up a tiny fraction of blue's space, has a fraction of the mass, generates a fraction of the heat, and requires less energy.  "Super"computer indeed...

Just as a point of reference, my machine is ~18 months old, and fairly well optimized and I have only reached peaks of 4.5 million operations per second in fritz, though I could probably eke out more, Blue Ice is worth nearly 100,000 of my PCs.

Forget "The Matrix" or "Terminator". Look at movies like "Enemy of the State" and "Eagle Eye” and the original computer takes over the world movie “Colossus, The Forbin Project”.

remember that flying on space was first shown on movies... right?

yahh i agree ... terminator is possible...

Very interesting, thanks. Deep Blue looks like the monolith in 2001: A Space Odyssey.