Ellon Musk artificial intelligence is one step closer to destroy the humanity... by playing Dota 2

If you still do not know, Ellon Musk has an artificial intelligence project called OpenAI under development.

I would not be the best person to go into the technical details of what is special about this artificial intelligence, but I know that the base of its development is to use a Machine Learning system so that the AI ​​learns by itself what it has to do.

Slightly scary. Machines... Learning...

But at the same time it is extremely fascinating.

Training the robots through games

Some of the experiments carried out to test the potential of this artifical intelligence can be found on this site. It is very interesting to see the robot learning to solve small problems through trial and error.

But OpenAI is already starting to stand out when it comes to solving more complex problems, and games, being essentially a set of problems to be solved, is being the main engine for this development.

Go, the simplest and most complex game that chess

Who haven´t heard of Deep Blue the IBM computer that defeated the chess master Garry Kasparov?

Computers / robots have been created quite successfully to be good chess players, but until recently, no computer had managed to be competent enough to win against a good Go player.

What makes Go more complex is the amount of possibilities that each move can bring, an amount the developed computers did not have the ability to process.

(Go is a very simple and interesting game but with an extreme level of strategy complexity. If you want to know how the game works, here is a guide.

At the end of last year, google's artificial intelligence, DeepMind was able to do it, after playing thousands and thousands of games against itself until learning the best strategies possible.

Dota 2, an even more complex game ...

![23x6gu4uxa.jpg] ()

Despite being mere 'computer games' filled with colorful characters, the e-sports scene has grown dramatically.

To get a sense of the size of this phenomenon, The International 7 (Last Dota 2 tournament) that occurred last year distributed a total of [24 Million dollars in prizes](http://dota2.prizetrac.kr/ international2017), being 10.8 Million for the champion team. Well... Computer games are no longer child's thing for some time now...

But how complex can it be?

Some Dota 2 numbers

The rules are simple: Two teams of 5 players on each side compete in the game arena to see who destroys the base of the other team first.

But the simplicity ends there ...

There are 115 heroes that can be chosen by each player, which generates a total of 14,189,367,287,524 possible combinations only in the choice of which heroes will enter the game.

In addition there are a lot of items that can be bought inside the game, changing the result of all possible interactions between the heroes of each team.

And this is only considering the gross amount of possibilities, because even a professional team has to consider in which way each player will stay, movement on the map, moment of attack, moment of defending, map control, vision of opponent's posiconamento, etc. etc etc etc ....

The first appearance of OpenAI

The first appearance of OpenAI in the professional landscape of Dota 2 appeared exactly in TI 7 (the one that distributed 24 million dollars).

At this event, OpenAI was set to play against professional players in a game mode that we will call here 'Shadow Fiend 1v1' (Shadow Fiend is a hero of the game), where two players use the same hero on the middle way and whoever kills the other 2 times wins the match.

It is a very simplified game, but it needs a certain specific amount of skill. The result?

All professional players (legends of the Dota, by the way), have lost.

You can check coverage here

The big question that remained was whether OpenAI could beat a team of pros in a traditional game of Dota 2 ...

OpenAI vs Ex-professionals

About a week ago, OpenAI showed up again, and this time to face a team of former professionals.

Spoilers: The professionals have lost badly.

A Dota game between two good teams takes on average 45 minutes to 1 hour and 30 minutes.

OpenAI won both games in just 25 minutes each.

The professionals won the 3rd game, but only because the heroes of the OpenAI team were chosen by the community through Twitch, who tried to choose the worst team possible.

Curious to see the match?

First, take a look at the announcement video of the match (which also explains some of the rules).

And here are the matches ...

** Game 1: **

** Game 2: **

** Game 3: **

Conclusion (or why is this event important)?

[Even Bill Gates was impressed by the result.](Https://www.cnbc.com/2018/06/27/bill-gates-openai-robots-beating-humans-at-dota-2-is-ai- milestone.html)

It is estimated that to become a professional Dota 2 player requires 12,000 to 15,000 hours of play, not counting other skills such as communication and teamwork.

In his training, OpenAI played 180 years in game time in just 1 day.

The technical details can be found here

But the most impressive thing of all this history is to begin to imagine the possibilities of the future when we have intelligent thefts capable of learning in minutes what we mere humans took years to learn.

And the possibilities range from the most wonderful to the most terrible ...

---

Want to follow my journey in the world of trade? Visit Trading passo a passo

---

Links worth a click

CoinTracking - Cryptos Portfolio Manager
Mannabase - Basic Income Token
freebitco.io - Best faucet I've found so far
Earn.com - Get Bitcoins to answer emails
Aelf - Earn Aelf tokens (ELF) doing daily tasks
Byteball - Register your steem username on byteball to recieve a big airdrop
Gods Unchained - Card game running on the blockchain