Chess, Computers and Humans

At one point in my life I thought it might be a good chess player. In the mid-80 collected and reviewed annotated games by Leontxo Garcia in El País -what big- and even I tried to make my first steps a little more seriously and played a federated year. Enough to realize that it would be better devoted to  write, but I always kept my love of a sport for years to continue sideways.

I did in those legendary duels between Karpov and Kasparov, but also with a nod to the passion that became part of my job: computer science. Expectantly read the news about the fighting between grandmasters and chess computers and cheated me over and over again thinking the man would always be higher than the machine. Absurd, as demonstrate a machine called Deep Blue the 11th of May 1997.

The beginning of the end

Leontxo Garcia wrote it better than anyone again in the country. In ” Man surrenders to the machine ” (curious, lack the tilde in the headline) this journalist recounted how Gari Kasparov ended up leaving the sixth game and losing by 3.5 to 2.5 against Deep Blue after an unexplainable mistake just initiated opening. IBM ride later much a tribute website as another to revisit the games that especially cool because web 1.0 is all.

Kasparov -leed this column his 1996 Time, please-, like the rest of those who attended that, do not believe. It could not have happened. There had to be trick. The legendary grandmaster accused the Deep Blue team of cheating and human intervention in the game Deep Blue, too “creative” at specific moments like that end of the first game (won by Kasparov) in which motion machine 44 he moved his rook for no apparent purpose.

The machine made ​​a move too sophisticated, one of which grandmaster Yasser Seirawan speak in a commentary for Wired in 2001 This publication Revisited starting time later, after the publication of the book. “The Signal and the Noise” by Nate Silver – attentive to the mini documentary that they would publish later in FiveThirtyEight- in which precisely appeared a unique theory: that the movement was simply a bug.

Far more humane Deep Blue was a mistake.

When he finished the game Kasparov asked to evaluate those items through records Deep Blue, but was not given. Those records were years later -today are accessible through the Web Archive (items 1, 2, 3, 4, 5, and 6) -, and those who later analyzed seem to conclude that those items and movements Deep Blue were legitimate.

What became clear is that in a way the team played not fully IBM fair, because in the tournament contract stipulated that could modify the program Deep Blue to polish their behavior in the middle of each game. They took advantage of that, but also rushed to dismantle Deep Blue after the games and that helped them an excuse not to give Kasparov a rematch.

That was the beginning of the end for the battle between humans and artificial intelligence machines.

The era of chess engines

Since then the history of chess grandmasters and clashes between programs and increasingly powerful computers made ​​it clear that this was a point of no return. Anand, Adams and other great masters fought against an increasingly polished and powerful artificial intelligence with mixed results, but Kramnik was the one who had the last really important game against a machine.

Would it have achieved the best Bobby Fisher of history beating Deep Blue?

It was in 2006 against Deep Fritz, a software that eventually won 4-2 (two defeats and four draws Kramnik) and sparking a comment particularly striking Professor Monty Newborn, one of the experts who helped organize the original match between Kasparov and Deep Blue. Speaking of future man-machine confrontation, this academic made ​​this clear: “the science is done.” The job was done and science had been imposed.

Since then no great master seemed interested in competing against machines. That simply had no emotion.

FIDE wanted to preserve the human game and did not admit that the machines might have an ELO with which to compare human players, and indeed the growing power of these programs made ​​them create side tournaments between machines competing with each other and in turn would gain certain points in a kind of ELO for machines managed by organizations such as the SSDF that eye, it is not directly comparable to human rating.

You may also like to read another article on TheKindle3Books: How long have you being the worst possible password 123456 and why never get delete it?

The list is continually updated and takes into account various factors, and is currently led by Stockfish 6 with a score of 3,334 points ELO “machine”. Magnus Carlsen, world champion and chess prodigy child, has today 2,844 points ELO officials.

Actually there are other lists that we find the classification of the different chess programs like LLCA the IPON or CEGT, and if there is a “world cup” of “engines” of chess -now called ASI- that’s the Thoresen Chess Competition Engines (TCEC) that its last edition was won by Komodo.

Could Magnus Carlsen beat machines?

Some wonder occasionally if it was possible a similar confrontation today and what would result. If Carlsen who many rivals compared to a machine for seamless game could pull off a victory against the machines. The answer is obvious to many: the machines would crush him.

He explained Carl Lumma, a software engineer who analyzed the question on Quora few months ago and began saying that first the ELO ratings of the machines had to be “rescaled” to make a comparison. Even using that score this programmer believed that a human player would if you have a perfect performance to 3,000 ELO points, while the chess engines exceed at least 200 points that score, which would mean an overwhelming advantage:

Summary: In a hypothetical tournament 12 games, Carlsen could hope to make a draw in 5 games and lose in the other 7 if you were optimally.

And the machines have not yet reached its optimum level. Some studies have estimated that an Elo rating of 3600 would put the chess engine and able to play the perfect game. With the current rate of gain ELO points of these engines -the arrival of Stockfish 6 ha made ​​to accelerate to 100 points / year- seems to indicate that this level will be reached by 2020. At that time chess machines will theoretically perfect.

The question, of course, is whether that makes a difference in some way. Years that even smartphones can beat grandmasters, but Anand said in the article, maybe that was depressing, but they do not even think it. “Since we’ve come,” he said, but not without adding a final Puntillita to autoconsolarse: “You can play that if you lose not survive.”

This future seemed clear and years ago, when Kramnik lost the duel, 2006. The odd thing is that in that article that year in The New York Times Newborn professor concluded by saying something unique:

If you are interested in programming computers to compete in games, the two most interesting are the poker and go. That’s where the action is.