Why did it take so long

[Dann Corbit]

Why did it take so long before a human world champion has been beaten by a computer.

It is clear that computers search much faster than a human. For instance if you have a large amount of data in a file and you want to find a certain pattern it is clear that computers can find it much faster.

Also they have much better memory.

So why didn't computers beat human world champions thirty years ago. Or is selective search something relatively new.

It took so long because the programmers were not good enough.

I think that the programmers were good enough (which is to say that they were every bit as talented as the programmers today.) The problem was that the filtering algorithms like LMR had not been invented yet.

With better software the top programs could beat the world champion earlier.

You mention the fact that humans are not perfect(search slower than chess programs and have worse memory) but humans are not perfect also means that they do not write the best chess programs so the world champion also has his relative advantage that is knowledge about chess that the opponent program does not know and not speed of calculation or memory.

Uri

The Deep Blue team was worried about your second point, so that had it worked into the contract that they were able to reprogram between games.

[tpoppins]

the modern era of man started with the industrial revolution approx. 1900's...

97 years...

not bad...

You're off by approx. 100 years.

[Dann Corbit]

the modern era of man started with the industrial revolution approx. 1900's...

97 years...

not bad...

You're off by approx. 100 years.

Abacus and Antikythera device: Thousands of years old

Arabic Numerals, Napiers Bones and Pascal's Pascaline: Many hundreds of years old

Acceleration of mathematics has been going on as long as humans have been trying to count things and otherwise do math.
https://en.wikipedia.org/wiki/Timeline_of_mathematics

[tpoppins]

@Dann: duly noted.

I suppose it boils down to the definition of "modern era of man". Most netizens these days would be hard-pressed to define that, as it would mean having to choose between the cell phone, YouTube or Facebook as the starting point.

[Uri Blass]

Why did it take so long before a human world champion has been beaten by a computer.

It is clear that computers search much faster than a human. For instance if you have a large amount of data in a file and you want to find a certain pattern it is clear that computers can find it much faster.

Also they have much better memory.

So why didn't computers beat human world champions thirty years ago. Or is selective search something relatively new.

It took so long because the programmers were not good enough.

I think that the programmers were good enough (which is to say that they were every bit as talented as the programmers today.) The problem was that the filtering algorithms like LMR had not been invented yet.

With better software the top programs could beat the world champion earlier.

You mention the fact that humans are not perfect(search slower than chess programs and have worse memory) but humans are not perfect also means that they do not write the best chess programs so the world champion also has his relative advantage that is knowledge about chess that the opponent program does not know and not speed of calculation or memory.

Uri

The Deep Blue team was worried about your second point, so that had it worked into the contract that they were able to reprogram between games.

I did not say that the programmers of today are better than the programmers of 1986 but only that the programmers were not good enough in order to beat the world champion at that time.

Thinking about the right algorithm and the right evaluation is part of being good enough and if programs could beat the world champion by using LMR and other tricks that programmers did not find at that time then it means that they were not good enough to do it.

I believe that it is possible to beat the world champion with hardware of 1986 with the right chess program(not sure if the knowledge that we have today is good enough for it).

Note that the reason that chess players lose against computers today is that chess players are not good enough and I do not compare chess players with chess players of the past but with the perfect chess player.

[duncan]

Why did it take so long before a human world champion has been beaten by a computer.

not something I know much about. I assume you need 14 ply search with a half decent evaluation to play world class.

you can either get there with fast hardware , clever algorithim or some combination.

chess is so big and the best algorithms are quite clever it just took a long time to get to 14 ply

[Jesse Gersenson]

Does anybody know what would be the oldest processor you need to beat human world champion chess. I mean no special chess chips.

Pentium III. Current linux compiles of Komodo and Stockfish failed when I tried running them on a Pentium I.

I'm testing Pentium II this week.

Pentium II was unable to run Stockfish. But, perhaps someone could hack together a version of which would run on the Pentium II. Any programmer interested in getting SF to run on the PII, I'd be willing to test the builds, pm me.

[yurikvelo]

Does anybody know what would be the oldest processor you need to beat human world champion chess. I mean no special chess chips.

it depends a lot on match rules

1) what kind of opening book is allowed, if at all
2) what time control for PC and for human, blitz, rapid, LTC, correspondence, Pondering = on/off
3) what EGTB is allowed, if at all
4) what are tournament conditions for human, 6 games in 6 days or 100 games in 25 days?
5) what are rules for machine team resign or offer draw. Is machine allowed to squize human player up to Mate (in hope for human to blunder) and to force opponent get tired for next game. Or machine team should be gentlemens and resign at -3 eval and offer draw if +-0.05 for 5 consecutive moves?
6) is human allowed to thoroughly examine this particular engine at particular HW behaviour for months before tournament start?

Depending on answers it may be somewhere between 486DX2 (if able to get SF or K9 compile for it) and dual-core AMD K8