In December 2017, I was UAE WhatsApp Number List  like many rare electronic chess enthusiasts, by the performance of AlphaZero, a program built by Deepmind, the Google subsidiary in charge of artificial intelligence. And moreover my partner Christian Semé in his 2018 greeting article recalled this important episode (first chapter of the article ). Let us briefly recall why this was a revolution two and a half years ago. DeepMind DeepMind A contrasting story, a mixture of developments and standing still The first chess programs date back to 1951. And as early as the 1950s, scientists had predicted the defeat of the best human player

against the machine in the early 1970s. This prognosis turned out to be false since, despite the rapid development technologies, hardware with Moore’s law, and software with advances in algorithms and programming languages, it was not until 1997 that the world chess champion Garry Kasparov lost his title against a computer ( DeepBluefrom IBM). Despite some progress on algorithms, the programming technique has generally remained the same from 1951 until 2017, i.e. almost 70 years! Digital technology is constantly evolving, of course, but breakthrough innovations seem to be on a whole different pace.

2 Years Later, In Covid-19 Confinement

For 70 years and it is not over yet, chess algorithms consisted of predicting each move and analyzing the resulting position to give it a score (for example between -100 and +100): a negative number meaning “advantageous position for blacks ”, and a positive number meaning“ advantageous position for whites ”. Then the program classifies all the moves, then then plans the next move, then the next move after, … Which results in a kind of tree of strokes and for each node of the tree a note which corresponds to the evaluation of the resulting position. The main algorithmic advances consisted in avoiding re-evaluating

UAE-Whatsapp-Number-List

irreparably bad branches of the tree and / or detecting branches that come together after a series of different strokes. In this programming method, what makes the big difference is the evaluation function. At the beginning, in 1951 for example, these evaluation functions were I imagine very very basic (roughly: if I have one more part I have a better position, if I have one less I have a less good position), then a little more “subtle”: the queen is worth 9, the knight is worth 5, … During the years 70-80-90 collaborations were set up between great chess masters and developers to program evaluation functions which were best inspired by

The New Territories Of Artificial Intelligence

the positional analyzes of the best human players. Until beating the world champion in 1997 and then overtaking human players, not because these programs evaluated positions better than human players (they could only come close) but simply because they calculated more than moves ahead because of their phenomenal computing power. As an example the best current program based on this technique – Until beating the world champion in 1997 and then overtaking human players, not because these programs evaluated positions better than human players (they could only come close) but simply because they

calculated more than moves ahead because of their phenomenal computing power. As an example the best current program based on this technique – Until beating the world champion in 1997 and then overtaking human players, not because these programs evaluated positions better than human players (they could only come close) but simply because they calculated more than moves ahead because of their phenomenal computing power. As an example the best current program based on this technique -Stockfish – rates nearly one hundred million shots per second! (millions of times more than human players). Please note that Stockfish is a completely free, open source program under GPLv3 license that runs on any PC, smartphone.

Leave a Reply

Your email address will not be published.