At the DOTA 2
world championship, The International
, a top contender has quite a different looking team — a set of five
artificial intelligence
robots.
OpenAI
, an artificial intelligence research firm that is funded by Tesla’s
Elon Musk
, has been taking on top Dota 2 players with its bots, in an exhibition match at one of the biggest events in eSports. [caption id=“attachment_4607051” align=“alignnone” width=“1280”] The International, DOTA2 Championships. Image: DOTA2[/caption] This is not the first time that the OpenAI is facing Dota 2 champions. It has previously defeated many leading players in the immensely popular MOBA
games. Also recently, it has won in matches against certain amateur teams. “Our team of five neural networks, OpenAI Five
, has started to defeat amateur human teams at Dota 2,” OpenAI mentions is their most recent post
on OpenAI blog. At any time, a character is able to make one of around 1,000 available actions; the bots have to make effective decisions while processing what’s going on in the game at a given point in time. To teach its bots what to do, OpenAI uses reinforcement learning methods, which works on a trial-and-error basis, and over time, the AI evolves. OpenAI Five, their team of 5 neural networks, plays 180 years worth of games against itself every day, learning via self-play. It is a large scale version of their earlier and simpler solo variant
of the game from last year. [caption id=“attachment_4607081” align=“alignnone” width=“1280”]
OpenAI Five playing the best OpenAI employee team. Image: OpenAI Blog[/caption] In just a couple of hours, the bots are able to learn more than a human can in their lifetime. However, machine learning works in a much different manner than humans, and that’s why the OpenAI bots are not yet the champions. OpenAI will be playing a top team on 28 July and the match can be watched in a live broadcast on
Twitch
. The post says that OpenAI’s goal is not to be the champions at the Dota 2 and win the largest share of the $15 million prize but to understand how training method can help AI solve all kinds of complex problems.