We have always been fascinated by the world of Artificial intelligence and slowly, what we have just seen in movies or science fiction books, is getting closer and closer to reality.

As is well known, some AI have already beaten champions of various disciplines as well chess or, more recently, they have burnt ground against the major teachers of the Go.

In defense of the "poor people" defeated by these programs, one has to say (especially how much it concerns Go) which are games in which mathematics has a major impact and for this reason an artificial intelligence has a great advantage over the human mind, thanks to its computing power (and therefore to be able to foresee practically every move available to the opponent and to respond to better).

On the other hand, the demonstration put in place during theInternational 2017 di Dota 2 this year, where artificial intelligence, called OpenAI, had the chance to show off against some of the top players in the circuit.

Usually, as surely you will know better than me, Dota puts a lot of emphasis on team play, but the game also allows a mode 1vs1 and it is precisely on this mode that the programmers of OpenAI they wanted to focus on showing their project against one of the most famous players, Danil "Dendi" Ishutin.

In this mode both players compete in Mid Lane and to win the win the only goals are either the destruction of the first tower or two kill. For this demonstration, some items (such as Bottle and Soul Ring) have been limited.

"Dendi"He chose to play Shadow Fiend, a hero cannon well suited for aggressive games, and has been making a mirror match against OpenAI.

The first move, really a "player" navigated, that he put in place OpenAI has been to make a maneuver of creep blocking, to force the opponent to move forward in his own territory (and therefore closer to his tower, to put him at risk of killing, also considering the fragility of the chosen hero).

Dendi, not reading the AI's move perfectly caught up in gambling, beginning to lose experience and gold, while OpenAI in the first three minutes, thanks to the greater number of last creep hits, he managed to attack the tower and earn the first kill.

Now the game is down in favor of OpenAI e Dendi, in a last desperate attempt to tear a kill, is facing the second death and delivering the first victory.

In rematch, things have gone even worse, with Dendi who undergoes the first kill even in the first two minutes of play while he was beating away.

OpenAI he is "trained" (according to the company's CTO Greg Brockman), playing virtually an immense number of matches (it's about an equivalent of hundreds of years of match). During the week he had already defeated other pro as well Sumaili e Arteezy, continuing his work of apprenticeship and improvement, considering that until then the bot had only played against itself, while second Greg Brockman, to complete the training it was necessary OpenAI playing against human opponents to learn how to react to various styles of play.

Behind the whole project OpenAI (funded in part by that Elon Musk) and the company that designed it is working on other AI projects, which range in various fields, but gaming is their favorite.


Now aware that they have achieved their purpose in the mode 1vs1 di Dota, which allows the bot to focus on a few elements that are part of the heart of the game (the strenuousness of a chess player who trains exclusively on the openings), the team of OpenAI has begun to design their new challenge, that is, to provide an effective and winning bot even in games 5vs5 by next year.

How will Skynet come from?