OpenAI's Dota bot almost becomes the champion of the game
99.4% from total domination...
Last week, Valve introduced a new arena mode for Dota 2 where players can challenge none other than the OpenAI Five bot. The bot is produced and trained by OpenAI, based in California. The Five bot made quite a name for itself when it managed to defeat a professional esports team at the game. The Five stands for its capability to handle five-on-five match format in the game.
But it gave a tough time to all who dared to try their might. Between the days from 18th April to 21st April, the bot faced 42,729 matches, combining cooperative and competitive. According to OpenAI, the bot bagged 4,075 victories which takes its victory rate up to 99.4%. This is significantly better than the human best victory rate of 24% in teams.
Moreover, ever since the launch of arena mode, it had been 459 games, until a human player finally bested the AI bot. The bot reigned supreme for almost six hours in the game.
OpenAI CTO Greg Brockman in expressing the bot's success said:
Arena was a massive-scale experiment to test whether OpenAI is exploitable, given the entire Internet trying to break it. The Dota community teamed up, cataloging every weakness. While Five has more to learn, no one was able to find the kinds of easy-to-execute exploits that human-programmed game bots suffer from.
He further stated how this success can help his company to achieve their goals:
This lets us have more confidence that future AI systems we deploy in the wild will be able to be made robust and hard to subvert. And perhaps even more importantly — we learned the value of having a community of people excited to pore over a system we’ve built in order to truly understand the limits and impacts of what we’ve built.
Ever since the bot's introduction in the arena, it has amassed an experience equivalent to 10.7 years worth of gaming. All this through facing human opponents. It also had some other commendable achievements. For instance, only 4 teams were able to beat it consecutively for 2 times. While 3 teams won 3 games and only one Dota 2 team was able to beat the beat consecutively for 10 times.
The OpenAI team aims to use this data to analyse what are the minute decisions that the bot makes. Moreover, what the the opportunistic plays the bot makes and what are the difference between the two approaches. OpenAI Five team researcher Jonathan Raiman said:
Arena was an immense source of anticipation and fear in the minds of many of us on the team. On one end of the spectrum we are absolutely sitting at the edge of our seat awaiting some incredible strategy to unfold, a missed blind spot pop up, or witness meaningful cooperation between AI and humans.
But there was another side to the story:
At the other end, our focus on the team had been to reach the highest levels of play at Dota 2 and hold our own against professional players, so it was a massive shift from our competitive mindset to the this world of investigation, external scrutiny, and robustness to Internet scale. I’m deeply thrilled and humbled by the public reaction. It’s a strong validation of what can be done with scaled up reinforcement learning, willingness to validate ideas in the real world, and a glimpse into what large scale AI deployments will be in the future.
Stay tuned for more.