OpenAI Bots Massacre Team Of Seasoned Dota 2 Veterans Live On Twitch

Dota 2
Researchers and scientists are doing some wonderful things with artificial intelligence (AI), from saving lives in healthcare to improving cybersecurity. AI also being used to beat up on professional gamers. A team of machine learning AI bots known as OpenAI Five once again had its way with professional Dota 2 players, this time on a full five-on-five match livestreamed on Twitch.

OpenAI Five is a team of five neural networks developed by OpenAI, an AI research lab founded by Elon Musk. While not quite the Skynet that Musk fears is a real possibility, OpenAI Five is adept at Dota 2, a game that requires a combination of high-level strategy and teamwork to be effective, especially against professional players.

In this case, OpenAI Five went up against some of the best Dota 2 players in the world. In two out of the three matches, OpenAI Five handily defeated its human opponents, and only lost a single match. The sole defeat only occurred after spectators were allowed to pick the playable characters, and in doing so they selected an imbalanced team.

On the flip side, OpenAI Five claimed victory after whittling down the list of heroes from over 100 to a pool of just 18. Even so, it was an impressive showing. OpenAI Five beat up some well known Dota 2 players, including William "Blitz" Lee and Ioannis "Fogged" Lucas, among others.
OpenAI Five has undergone extensive training leading up to this point. It plays 180 years worth of games against itself every single day, learning strategies via self-play. It trains using a scaled-up version of Proximal Policy Optimiztion running on 256 GPUs and 128,000 CPU cores.

While beating up human opponents in Dota 2 is neat, there are bigger goals that OpenAI is working towards.

"Our underlying motivation reaches beyond Dota. Real-world AI deployments will need to deal with the challenges raised by Dota which are not reflected in Chess, Go, Atari games, or Mujoco benchmark tasks," the OpenAI team adds. "Ultimately, we will measure the success of our Dota system in its application to real-world tasks."