AlphaZero is already better at various board games than humans

AlphaZero

We have known for quite some time that one of the divisions of A, specifically the one baptized with the name of DeepMind, which is in charge of the development of the different projects related to the world of artificial intelligence within the North American company, works on the development of software that is capable of beating any human contender at various table games.

Specifically I want to talk to you about the software AlphaZero, which we talked about for a long time and that after many months, in which it has continued to develop, has managed to improve to such an extent that today it has already proven to be the best player in the world in practically all the games it knows . The best of all this, or at least this is what DeepMind assures us is that AlphaZero is just training.

go

AlphaZero is already the best player in the world in several of the most complex board games created by man

As you will surely remember, a few months ago the engineers in charge of the development of AlphaZero already made their project much better than any human in different board games. After all this time, apparently, those responsible decided to add big improvements to your artificial intelligence software So this new version will face the previous one. The results were impressive, after several hours, this version was already the best in the world.

With such a capacity for improvement, it is not surprising that its creators have decided to take its capacity to other board games such as chess or shogi, where it has already managed to be the best in the world in both although, personally I have to admit that it What is most striking is the way in which this software 'learn', since those responsible they just show her the rules of the game and let her playIn other words, this software does not seek to be the best in the world, it is only training.

chess

AlphaZero is capable of training itself

This is precisely what we can extract from the last paper published by those responsible for the development of AlphaZero where it is commented that after a rather expensive development in terms of code implementation and test tests they got their capacities to grow exponentially. An example of everything is that, for AlphaZero to learn to play Go, they only added the game rules and made it play against the version that was already the best in the world ... after just a few hours AlphaZero had managed to win by 100 wins to 0.

Extrapolating this to other board games we find out that something similar has happened, an example we have in the chess where, just by knowing the rules and after a training of only 4 hours, AlphaZero was able to beat none other than Stockfish, one of the most powerful chess engines in the world. We have a new example in the shogi, a kind of game very similar to chess but of Japanese origin where, with just two hours of training has managed to be unbeatable.

shogi

The purpose that DeepMind has for this software is to make it learn anything by itself

Surely by now you will have realized that AlphaZero is finally becoming an expert in board games although, the truth is that the engineers and scientists behind the project do not seek this end but rather their goal is much greater, manage to extract their learning techniques to implement them in many other areas, that is, they seek to achieve an algorithm capable of learning anything, something very similar to what happens with humans.

Although it may seem that there is still a long way to go, something that is true, we must take into account the great progress they are making in DeepMind with their artificial intelligence engines, same as evolve and refine at a constant rate so surely and sooner than we can imagine, we will finally be faced with an artificial intelligence capable of learning anything, task, work ... by itself and without the need for explanations.

Further information: MIT


The content of the article adheres to our principles of editorial ethics. To report an error click here!.

Be the first to comment

Leave a Comment

Your email address will not be published. Required fields are marked with *

*

*

  1. Responsible for the data: Miguel Ángel Gatón
  2. Purpose of the data: Control SPAM, comment management.
  3. Legitimation: Your consent
  4. Communication of the data: The data will not be communicated to third parties except by legal obligation.
  5. Data storage: Database hosted by Occentus Networks (EU)
  6. Rights: At any time you can limit, recover and delete your information.