Friday at The International, Valve’s big eSports competition in Seattle, the nonprofit OpenAI artificial intelligence research company unveiled a bot that defeated one of the world’s top professional gamers in the video game Dota 2, twice, live at Key Arena. As shown in the video below, the bot also trounced other top gamers in advance of the live event.
Elon Musk (who helps to fund Open AI) is calling this a “vastly more complex” achievement than a computer defeating a human in Go or chess, as reported by GeekWire’s Alan Boyle in our coverage of the demonstration.
This is not like playing a game against a computer in the way we knew it as kids. Open AI’s bot is much smarter. It was programmed to learn the game from scratch, playing against a copy of itself over and over, figuring out which strategies and tricks work to win the game, and then applying that knowledge to its gameplay.
Based on the descriptions from OpenAI team, the bot doesn’t benefit from any special system-level knowledge of the other player’s moves. It learns and plays the same way a human would learn, but on an exponentially larger scale.
“It’s played for really lifetimes of experience,” said Greg Brotman, co-founder and CTO at OpenAI, after the bot won its first live match on Friday. “It’s played so many games of Dota. It’s explored many different strategies. Learned to bait. Learned to exploit other people who bait. It’s just played far more into the strategy space than any human has.”
OpenAI says on its site, “The bot learned the game from scratch by self-play, and does not use imitation learning or tree search. This is a step towards building AI systems which accomplish well-defined goals in messy, complicated situations involving real humans.”
The post adds, “In the above video you can see that our bot has learned — entirely via self-play — to predict where other players will move, to improvise in response to unfamiliar situations, and how to influence the other player’s allied units to help it succeed.”
This is about a lot more than video games. Musk has warned repeatedly about the dangers of AI. The SpaceX and Tesla founder reiterated his concerns — comparing the risk to the current North Korean nuclear tensions — after the Open AI bot trounced 27-year-old pro gamer Danylo “Dendi” Ishutin in Dota 2.
If you're not concerned about AI safety, you should be. Vastly more risk than North Korea. pic.twitter.com/2z0tiid0lc
— Elon Musk (@elonmusk) August 12, 2017
The stated purpose of Open AI is to ensure artificial intelligence is created “in the way that is most likely to benefit humanity as a whole.”
Musk, Stephen Hawking and others are so concerned about the potential for autonomous weapons systems, and calling for a ban. To understand why, watch the bot play Dota 2 and then imagine if AI controlled a real-world weapons system. That’s not a game anyone would want to play.