AI Obliterates Semi-Professional Dota 2 Players In A Game

Adjust Comment Print

Valve's multiplayer online battle arena video game, Dota 2, is the latest example of that.

OpenAI, which has been in development as a Dota 2 bot since The International 7 past year, went up against a full team of five highly-skilled human players-composed of Ioannis "Fogged" Loucas, William "Blitz" Lee, Ben "Merlini" Wu, David "MoonMeander" Tan, and Austin "Capitalist" Walsh. In the third game, human players secured a win. To everyone's awe, the bots beat the human players, which included former Dota 2 professionals, 2-1 in a set of three games.

And the OpenAI has the advantage of learning through rigorous reinforcement training, a machine learning technique that involves teaching AI agents through trial and error and then "rewarding" them when they carry out a good action in the game. OpenAI's efforts aren't done yet as they are expected to take the bots to Valve's The International DotA 2 tournament to compete against some of the world's best teams.

While OpenAI's defeat of some of Dota 2's cream of the crop players is decidedly impressive, what's definitely more noteworthy is the unprecedented advancement of artificial intelligence itself.

After losing the first three games of Go to Google's AlphaGo, Lee Sedol, a professional player ranked second in worldwide titles, deployed a highly unusual strategy to win the fourth game, inspired in part by the idiosyncratic algorithmic playing style of the AI that he had observed during the previous three games.

Argentina's Senate rejects bill to legalize abortion
Catholic and evangelical groups protested abortion with the slogan, "Argentina, filicide (child murder) will be your ruin". However, activists also say that the fact that the bill was even debated in the first place was a "victory" of sorts.

You might remember OpenAI popping up in the news just recently. "You have imperfect information, you have team work, you have these exponential combinations of different heroes and items, and you have to be able to deal with all of that".

Following the victory, Musk tweeted congratulating the OpenAI team. It won the first game in 21 minutes and 37 seconds.

The tournament was a best of three but clearly designed for fun more than goal.

This OpenAI video is meant to provide some insight into the model's planning via an output which predicts where a hero will be in the future. Despite the wealth of experience on the team, the humans could do little against the never-ending onslaught and precision of OpenAI, and they quickly lost in humiliating fashion. In terms of compute time the Auguest 5th model uses 190 petaflop/s-days, which is a lot of number crunching.