Google's AI secretly trains in online games

[G.O.G]

Google is not content with having one of the most advanced artificial intelligence in the world. He wants more, and for this he needs to perfect the software with real field tests. AlphaGo, which is what it is called, already made an exhibition by defeating the Go world champion. Since then he has been training in secret.

To sharpen the performance of your Artificial Intelligence, Google has put online Go, a game very po[CENSORED]r in Asia. It has done so in secret, under an identity that, ironically, they have baptized as Master. Several players suspected that they were facing a computer to see this fictitious opponent win 50 of the 51 matches played.

Now Google has confirmed it: Master was neither more nor less than an AlphaGo, a program created by DeepMind, the company's subsidiary specialized artificial interface. During the weeks that he has faced other Go players online, the company has been able to fix some of its software flaws, making AI even more effective and powerful.

It was Demis Hassabis, founder of DeepMind, who has confirmed all the rumors in a tweet. In the statement also leaves the door open for the official participation of its AI interface in some official tournament: "Now that our unofficial tests are over, we are open to playing some official and complete game this year."

The possibilities of artificial intelligence are completely open, so much so that it starts to scare some. And if not ask the Japanese workers who are being replaced by robots. At the moment, do not be alarmed, at least not as long as it does not become a danger, which could happen.

The IA lacks years of development to become a really useful invention on a large scale. For the moment, beyond testing and exhibits, its practical application is really limited.