Tuesday , September 22 2020
Home / chile / "AlphaZero overcomes us deeply": the computer that generates its own knowledge and plays as a "superhuman"

"AlphaZero overcomes us deeply": the computer that generates its own knowledge and plays as a "superhuman"


robotRights of the image author

Image caption

There are machines that behave like Superhuman.

For some time now, machines have been a challenge to overcome humans.

In many respects, robots have proven to be "Harder, better, faster, stronger", As the Daft Punk song says.

Now, your challenge is to be more elegant, Or at least that is what AlphaZero computer audiences, one of DeepMind's most recent developments, the Google company dedicated to the artificial intelligence

One year ago, AlphaZero beat out forcefully to other machines trained to play chess; shogi, known as Japanese chess; and go, one of the most complex board games that exist.

  • Magnus Carlsen, world chess champion: who is the Norwegian who fashioned this sport

Now, a year later, the creators of AlphaZero reveal the keys to how the machine smashed To their opponents, playing at a "superhuman" level, according to their creators.

Rights of the image author

Image caption

At the shogi, I consider Japanese chess.

"AlphaZero learned to play three challenging levels games taller which they have seen, "he wrote in the magazine Science Murray Campbell, IBM Researcher.

It is not necessary to mention that it overcame the humans, since this barrier was surpassed 20 years ago, when the Deep Blue computer overcame to Garry Kasparov, Perhaps the best sequel for all time.

  • What we can learn from the feminine prodigies of chess, a sport dominated by men

But the machines, although more powerful, until now fail to emulate human characteristics such as the instinct or the imagination

"As airplanes do not damage their wings like birds, machines do not generate chess movements such as humans"Kasparov wrote in a recent article in Science.

But that seems to be changing with the arrival of AlphaZero.

Rights of the image author

Image caption

Gary Kasparov was "pleased" with AlphaZero's way of playing.


In the Armenian squad Tigran Petrosian, who was the world champion, the phrase is attributed to him: "the best coach Chess is the same ".

And that was just what AlphaZero did.

  • Why do Armenians triumph over chess?

Until now, the most powerful artificial intelligence programs are based on a combination of search techniques, adaptation to the environment that is presented and features that allow them evaluate situations.

All of this, obviously, comes from human beings that are train And they schedule them according to what they consider to be the most optimal.

But that was not the case with AlphaZero, who was simply given the rules of the game and from here it strengthened this learning by playing against herself.

Rights of the image author

Image caption

AlphaZero was his own chess coach.

In a few hours, AlphaZero played more games against herself than those that were registered in the entire story of chess

"The programs usually reflect the priorities and the prejudices of programmers, "wrote Kasparov." But as AlphaZero program itself, I would say that his style reflects the truth. "

Thus, AlphaZero is distant from the way a human would play and begins to show ways amazing do not understand chess.

In his article, Kasparov admits that he was "pleased" to see AlphaZero play with a "dynamic and open" style, preferring positions that for a human like him would seem "Risky and aggressive."

According to Kasparov, he could observe that AlphaZero "re-evaluate Fundamental concepts such as the value of the pieces "and" prioritize the activity of the pieces on the material. "

Rights of the image author

Image caption

Artificial intelligence has applications in many fields.

Part of the success of AlphaZero is that it is not simply based on its ability to analyze thousands of possibilities in a second, as a simple calculator, But focuses your search on those games that look more winners.

  • The "Turkish" trick that destroyed the reputation of expert chess players in the eighteenth-century Europe

In other words, AlphaZero does not work harder, if not in more ways intelligent.

Although for now AlphaZero has no other role than playing chess, shogi and go, its creators hope that this technology can be applied to other areas such as science and technology. medicine

"AlphaZero is overtaking us a deep and useful way," says Kaspárov, "a model that can double in any other task or field where it can be generated knowledge virtual ".

This is what DeepMind says, which states that its objective is "to develop programs that can learn to solve any complex problem without needing to be teach how to do it ".

You can now receive notifications from BBC News World. Download the new version of our app and activate them so as not to lose our best content.

  • Do you know our YouTube channel? Subscribe!
Source link