Please forgive me my bad English and quick typing. Here it is:Vinvin wrote:29 minutes long : http://www.youtube.com/watch?v=4ttcO4UU9RU
In German. It would be cool if someone would translate in English
WCCC Cologne (1986)
0
Titel: Chess programs want to be world champion
background Vlastimil Hort: There white wins with a pawn more in rook endgame.
do not laugh, this is a ...
Titel: computer chess championship 1986 in cologne
Albrecht Fölsing: moderator, greetings
The tournament comes to the end, the games are in the 5th hour.
1
a strong group of 4 programs, more at the end
20 programs are here, most on mainframes over the world,
usa, canada, germany, but also small chess computers that you can buy.
in the 60th when chess programming started there were money
even from military sources, to improve strategy, but expectations not fulfilled.
chess programming is innocent again.
commercial small computers, hobby programmers
2
trying to build a machine that beats human world champion.
even strong programs here are still not strong enough for grandmaster play.
Why even the biggest computers still have problems shows the 1st film:
In our century the thought of a chess playing machine is still fascinating.
In the french novel "check mate the roboter" such a machine is the main character.
The machine in human form plays excellent chess.
3
But the style makes people suspicious. It plays like "Robert"
with the same moves as 2 years before. Robert was murdered
and his brain is in the machine.
No SF was a chess game between David Levy in Hamburg
and the best chess machine in USA with a robot arm.
4
Levy had made a bet 1969 that in the next 10 years no computer
could beat him at chess. After 6 hours the game with winning chances
on both sides ended remis.
Relaxing chess players with simpler backgammon where chance
5
and strategy are equal important. Even a small computer can play backgammon.
Chess is more difficult. The strongest chess programs need the power
of the biggest computers of the world to play reasonable chess.
This 20 million DM computer can do 80 million results a second.
In 1 second more tha a mathematican in a whole life.
Made of chips and integrated circuits.
6
Top chess programs need that for the brute force method
where all moves are done up to a fixed depth.
The computer does not distinguish between reasonal and silly moves.
It generates 100 thousends end positions that it evaluates.
The better the computer the deeper the search.
Such blind search has its limits.
The number of positions on the board is unlimited.
In a typical position each player has 35 possible moves.
Symbolized by 35 pieces of sand. then 40000,
7
more than a million, 50 million a bucket of sand,
4th move of white 64 billion positions, a truck full o sand.
A computer has to handle this in 3 minutes.
8
Only the world's biggest computer can do this
and there is obviously a natural limit because soon
the sand of the whole area represents the positions.
For games with 50 moves there are more variants than atoms in the known universe.
Moderator: That is a very impressive demonstration
why the biggest computers have a hard time to find a way through the chess universe.
9
Since we began to teach computers to play chess 30 years ago
there were two strategies.
One was the implementation of human grandmasters knowledge but
that is not very successful.
The other method is to let the computer as a busy idiot
do all positions and find the best move by brute force.
That way there is some power and success.
This tournament has already had some improvements.
Even the small computers are good.
The first example is Enterprise that you can soon buy for 300DM.
From Carl Danielson, 23, from Denmark. 1 year development.
10
Brute force or chess knowledge? Much knowledge in a brute force program.
Impressive play, you won against?
Won against an Amdahl, 10 million DM, from Canada.
30 s time on Amdahl costs the same as the small computer.
Good luck.
11
Similar here with a program on the mainframe of TH Braunschweig.
Against the Mephisto computer. Developer is Mr. Schrüfer,
what was the tournament for you?
We could beat the world master program with some luck,
lost the next one and this one is very sharp.
Ossi Weiner you helped develop Mephisto, how good
do you think are the small computers.
They have made progress in the last years.
And we are proud to be among the top teams. We have hope.
12
It is questionable if you (at TV) could beat the small computers
because they are so powerful.
Therre is a second tournament for everybody against computers
10 minutes time for the machines, 60 minutes for humans.
Club player against computer: The difference is the human
makes errors the computer not but the human has more ideas.
13
In this game there are small advantages for black/human.
Moderator: It needs club players to win against the computer
and who wins may barticipate at another tournament with prices,
chess computers of course.
14
There is another expert, the writer (satyricer) Ephraim Kishon.
How good do you play? EK: How bad i play. I was good 40 years ago.
Computer chess is your passion How many computers do you have?
11 now but I gave half of my computers away.
Do you play against them?
15
Not really. I analyse a lot while I am writing.
Writing alone is boring.
What says the best wife of all?
She believes I am crazy. Threatens to leave me but of course is still at home.
The best players spend a lot of time with chess
and have capabilities to beat the best computers.
16
How can humans handle the labyrinth of chess positions?
In a blitz tournament (10 minutes) there are nice games.
Masters sometimes make the kibitzing crowd cheer.
German blitz master Bernd Fäustel plays a tricky stalemate.
The grand masters have the ability to look at a game and
understand it very fast.
17
There is an impressive demonstration in an experiment.
After 5 seconds the hungarian grand master Androsch Adorjan
one of the 10 best players worldwide, rebuilds a complicated
position. And he is disappointed when a rook was placed good but wrong.
An amateur or novice cannot do it. He is lost.
18
Typical they can do only 7 and this one only 1 correct piece.
The legendary abilities of a grand master are not based on a perfecr memory.
Here is an experiment with a random position.
19
The grand master does not like it. I give up, this is hopeless.
Androsch Adorjan can do 5 pieces.
These experiments show that chess players do not calculate
but handle a big repertoire of piece patterns.
With their chess vocabulary grand master have a different view to
positions than a amateur. The important things are seen immediately.
20
We use a technic from customer behaviour research from University Saarbrücken.
There is a tracking of eye movement on advertisements.
Here the man did not look at the precious jewel in the picture.
Target missed.
Another test. First a novice. His eyes move around without plan.
21
From piece to piece but he does not see the combinations.
With grandmaster Adorjan that is a different story.
He looks at the opponent king in a position he does not know.
This is really easy hee says and plays the mate combination.
22
He looks at the squares that are influenced by the pieces and the king cannot escape.
Moderator: The experiments you have seen are based on a book
written 40 years ago by psychologist DeGroth from the Netherlands
Who is an excellent chess player. What do you think is the
difference between man and computer?
23
You have to calculate and evaluate and decide.
Computers can do that fast but men can see the whole picture.
The computer will never learn the human intuition.
Will a computer beat a human grandmaster?
I don't know but people underestimate the difficulty.
24
Computers will not become world champion in the year 200 or earlier.
Moderator: Dr. Berliner you are professor of computer science in america
and you have a special chess machine in Pittsburg and that is
a a winning candidate here. How big is it?
It is as big as 4 shoe boxes. The special machine cost 15000$
and a connected computer costs 25000$. It can do 30 million
positions in 3 minutes.
25
The niveau compared to human chess is 2350 in USA that is 2250 in Europe.
That is german Bundesliga (master class).
We expect an improvement of 100 points this year and 100 points the next year.
Then it will not be long and only a few people can hold that strength.
Hopefully yes.
You have been mail chess world champion. Do you think computers
will beat the human world champion?
26
Yes I think in 5 years. That we want to see. Good luck.
Dear viewer predictions a hard especially concerning the future.
Dr Berliners program Hitech had won 4 games but today he will
probably lose to the old champion Cray Blitz.
27
And that was my prediction: 4 programs with 4 points out of 5 games:
The old champion Cray Blitz, Hitech, Rebel an experimental micro
processor that always comes with a little fan to blow the heat away.
It was good against the other programs and you may be able to
buy it soon.
28
Estimations and extrapolations are that computers can beat man in
the time between 1995 and 2075, a very long time span.
Whenever that will be it is probably a silly brute force machine.
But even then computers still cannot simulate the typical human abilities.
This is how I want to say goodbye.
29
Harald
