AI and chess. An unstoppable evolution, from that “Kempelen automaton” to the beating of AlphaZero

As a kind of prologue to a retrospective look at the evolution of artificial intelligence in chessone could mention the story of an alleged machine, which was a great success in its time: el automata de Kempelen. Back in 1780, at the beginning of the Industrial Revolution, a Hungarian inventor known by the fictitious name of Baron Kempelen, presented in the main cities of Europe, a machine that played chess and defeated almost all the opponents who dared to play. face it. A sensation and a mystery. Certainly, at a time when the bicycle did not even exist, there could hardly have been a computer that played chess.

Machine It was a piece of furniture that hid inside an expert player, out of sight of the public, and using a mannequin to make the movements on the board. Over time this device became known, and the automaton fell into oblivion. A fraud perhaps, but also an early hint at the aspiration of human beings to create an intelligent machine that can think for itself, and in this case, play chess.

A reproduction of Kempelen’s chess-playing automaton; On the right, the cushion for the hidden player’s comfortWikipedia

The first real computers that played chess were manufactured in the seventies of the last century. At first they were hulks that barely reached an incipient level of play. In the eighties, I remember that a computer that played chess was marketed with the name of Mephisto; His level of play was basic, but enough to beat a beginner amateur. Then, In the nineties, the gaming power of the machines grew exponentially. It began to become common for a computer to play masters tournaments managed by an operator. The matrix system allowed it to calculate millions of moves per second, which translated into enormous practical power.

This system had a weakness; the computer’s position evaluation formula was getting stuck on closed positions. To be graphic, in a very blocked position, without movement of pawns, where a master could quickly conceive an adequate plan, based on general ideas, the computer instead continued calculating millions of moves, and could sometimes make a mistake, based for example in a materialistic evaluation of the position. Therefore, most teachers tried to induce the computer to play closed positions.

The Russian Vladimir Kramnik, in action in a series of games against the Deep Fritz program in Bonn, GermanyHERMANN J. KNIPPERTZ – AP

But this weakness was relative; the brute force of computer calculation prevailed most of the time. Thus, the defense of the dignity of the human being was left in the hands of the then world champion, Garry Kasparov. But not even a modest objective, such as preventing artificial intelligence from surpassing human intelligence before the end of the twentieth century, could be achieved. In 1997, in New York, the Deep Blue program He defeated the world champion in a six-game match by 3 ½ to 2 ½.

When the last game was reached, the match was tied. Kasparov, with black pieces, chose an opening, the Caro Kann, which at that time did not appear in his repertoire. He did this because in the main variant, the most fearsome option in favor of White was a knight sacrifice, and computers, at that time, did not sacrifice material, unless the aforementioned sacrifice provided a definite advantage, which was not. the case of this opening, at least in appearance. Kasparov speculated that if White did not make the sacrifice, the position would be satisfactory for Black. When the critical moment arrived, to the surprise of many, the compu sacrificed the horse and crushed Kasparov in just a few moves. The champion was left clutching his head, in an image that became iconic.

There was still the feeling that Kasparov could have played better, and for some years the best chess players in the world, especially Kramnik and Anand, kept their finger on the pulse of the machines. But before the first decade of the new century was over, the level of machine play skyrocketed again, and it was now impossible for any human chess player to mount any serious resistance.

Deep Blue, the machine that defeated Garry Kasparov

A new chapter to this story was added in 2017 when, a new computer development, the program called AlphaZero, equipped with a novel learning algorithm, learned to play chess one day in the morning, played thousands of games against itself, and in the evening of the same day it beat the best chess computers of the time. She made a hundred-game match between AlphaZero and Stockfish, which was the best chess program at the time, and AlphaZero won 28 games and drew the remaining 72, without losing any. And since then these computers have continued to improve their level of chess. To make a comparison, the best player in the world, Magnus Carlsen, has an Elo ranking of 2830 points, and the 20th player in the world, about a hundred points less; while the best computers exceed 3500 ranking points. Carlsen once commented that competing against a computer could be devastating to a chess player’s ego.

If you can’t handle them…

The truth is that computing and chess combine perfectly. Every day hundreds of thousands of chess games are played online through the many platforms that exist for this use. No teacher can do without studying without a computer assistant. The work tool of a great teacher today is a computer; With it he studies, prepares openings, rehearses variants, and learns from the machine’s combinations. Teachers no longer play against computers, but with them.

Pablo Ricardi

Themes

Chess StoriesGarry Kasparov

Other news from Chess Stories

What few know. The Story of Howard Staunton: How He Boosted the Game and Why Chess Pieces Are Named After Him

960 o Fischer Random. A new type of chess becomes popular in the world: what it consists of and what problem it faces

“Chess is my life.” The different player who burst in like a meteor, became Argentine champion and seeks to be a great teacher