OpenAI Five Defeated Professional Champion Dota 2 Team, Twice

14/04/2019

Dota 2 is a multiplayer online battle arena video game developed and published by Valve Corporation. Due to its complexity, the game has been used by AI companies to test and showcase their AI agents' capabilities.

This includes OpenAI, the AI research organization co-founded by Elon Musk.

OpenAI has previously defeated the world's best human Dota 2 players. And here, it goes beyond that by defeating a world champion e-sports team.

The competition held in San Francisco was dubbed as the OpenAI Five Finals, and it ended with the organization’s public demonstrations of its Dota-playing technology on a high note.

The competition on the human side included five top Dota 2 professionals from team OG, which won the world’s most coveted e-sports prize in 2018 when it took the number one spot at Valve's 2018 International, in a best-of-three exhibition match.

OpenAI's Dota 2 battle arena against human professional players
OpenAI's Dota 2 battle arena against human professional players

Here, OG competed in a best-of-three contest against the OpenAI Five bots, which were all trained using the same deep reinforcement learning techniques and controlled independently by different layers of the same system previously showcased by the organization.

After previously narrowly lost two games to less capable e-sport team in August 2018, this time, OpenAI's performance at the event was considered at its peak to date.

The deep learning system thrived by relying on aggressive and unconventional methods, reliance on a number of aggressive tactics, including instant revivals for heroes in the early stages, and picking only valuable heroes.

While OG put up a fight, the AI won the first two matches of three in a Captain’s Draft mode, which allows each team to ban characters to prevent the other from selecting them.

In the second match, OpenAI Five also emerged victorious after about 20 minutes.

OpenAI which is backed by the likes of Reid Hoffman and Peter Thiel, have created numerous AI agents for different purposes, including natural language systems capable of impressive coherency, and those capable of playing games like Pong, Montezuma’s Revenge, and Dota 2.

Dota 2 here is considered a vastly complex strategy game, involving more than 100 unique characters, deep skill trees, a huge item lists, and an array of variables at any given moment in a match. This makes it a perfect test bed for AIs.

For OpenAI, the victory is not only for a milestone celebration. It is also a testament that its approach to reinforcement learning and its philosophy about AI and intelligent machines.

The research team will no longer hold any public demonstrations of its AI agents, and it's working on software that will let humans collaborate alongside the OpenAI Five in real time, playing on a team with the bots and learning from their peculiar, unprecedented strategies and behaviors.

According to Sam Altman, the co-founder and CEO of OpenAI, the organization will likely continue to dabble with Dota 2 and other video game environments, primarily because they are such good test beds for AI and good benchmarking tools for measuring progress.

Ultimately, OpenAI wants to take its Dota 2 learnings and expand them to new domains outside games and, eventually, into the real world.

“What OpenAI is trying to do is build general artificial intelligence and to share those benefits with the world and make sure it’s safe,” said Altman, referring to the organization's quest to build a multi-purpose AI system capable of performing any task a human can.

“Were not here to beat video games, as fun as that is. We’re here to uncover secrets along the path the AGI.”