The machine has been a bit more than a challenge to overcome human beings.
In many ways, robots have proven "harder, better, faster and stronger"", As Daft Punken's song says.
Now, the challenge is more elegant, or at least that's what AlphaZero shows, one of the most recent developments in DeepMind, artificial intelligence
One year ago, AlphaZero overwhelmed other machines trained primarily chess; shogi, known as the Japanese chess; and go, one of the most complex complex games that exist.
- Magnus Carlsen, world chess champion: who is this Norwegian fashion game
Now, a year later, AlphaZero creators reveal the keys of each key crumble his opponents, playing on "superhuman" level, according to their creators.
"AlphaCerio played three challenges at all levels taller that's what happened, "he wrote in the journal Science Murray Campbell, IBM Researcher.
It is not necessary to mention that overcoming a man's obsession over 20 years ago when he won the Deep Blue computer Garry KasparovPerhaps the best chess player of all time.
- What can we learn from the female feminine chess, a sport that dominates men?
Machines, even stronger ones, do not produce human characteristics until now, such as instinct or imagination
"Because the planes do not have wings like birds, the machines do not generate chess movements human"Kasparov wrote in a new article Science.
But that seems changing Along with AlphaZero.
Tigran Petrosian, a charming Armenian, was the world champion, and "is best" coach chess is one of ones ".
And that's what AlphaZero was.
- Why do Armenians get chess?
Until now, the most powerful artificial intelligence program is based on the combination of search techniques to adapt them to the environment they present to them and provide these functions. evaluate situations.
All this, of course, comes from man train and they schedule them according to what is most optimal.
But this was not the case with AlphaZero, because he was saying the rules of the game and that's why he was learning to play. against him
Within a few hours, AlphaZero has played more matches against him, than all those who have registered history chess
"The program reflects priorities and priorities prejudices Programmers, "Kasparov wrote." But as the AlphaZero program says, its style reflects the truth. "
In this way, AlphaZero moves away to show a way to play human and way amazing to understand chess
According to Kasparov, he acknowledged that he was "happy" to see AlphaZero's "dynamic and open" style, and it seems like such a man "Dangerous and aggressive".
Kasparov says that AlphaZero "reversing Key concepts, for example, the value of the pieces "and" the default activity of the material parts ".
One part of AlphaZero's success is not enough but the ability to analyze a second option, but rather calculator, but searches for performances that bring more winners.
- The "eighteenth century Europe" chess experts destroyed the reputation of "Turkey" trick
That is, AlphaZero is no harder, no more smart.
For the moment, AlphaZero also has a function that plays chess, shogi and go, whose creators hope that this technology can be applied in other fields like science and science. medicine
"AlphaZero overcomes us in a profound and useful way," explains Kasparov, "a model that can be duplicated in any other task or scope. knowledge virtual ".
That is, DeepMind says its goal is to "solve problems with developing programs that are capable of learning from any complex problem. show how to do it "
Now you can receive BBC News World notifications. Download the new version of our app and activate them not to miss our best content.
- Do you already know our YouTube channel? Subscriber!