Elon Musk's OpenAI Just Beat The World's Most Unpredictable 'DotA' Player at His Own Game

Monday, 14 August 2017 - 10:49AM
Technology
Artificial Intelligence
Monday, 14 August 2017 - 10:49AM
Elon Musk's OpenAI Just Beat The World's Most Unpredictable 'DotA' Player at His Own Game
Image credit: NoobFromUA (YouTube), DotA2, The International
It's been twenty years since Deep Blue beat Garry Kasparov, the world chess champion, at his own game. Since then, AI has mastered Go and even Ms. Pacman. Now artificial intelligence has reached a new milestone with the news that Elon Musk's OpenAI just conquered the field of eSports, taking to the main stage of the digital tournament The International and wiping the floor with the best Dota 2 players on Earth.

Elon Musk himself is understandably very proud of this achievement.




What makes eSports, especially DotA, so unique, is that players can succeed not just by learning meta-strategies, but by "breaking" the game—finding loopholes and exploits that allow them to pull off otherwise impossible maneuvers. Dendi, one of the players that OpenAI was able to beat, made DotA 2 history by beating another world-class team at The International almost single-handedly by using an exploit called a "fountain hook," which allowed his team to turn around an otherwise doomed match. His trademark tactic in matches has long been to ">deliberately remain unpredictable, thus rendering all of his opponents' practice and training useless as they come across someone whose movements they can't comfortably predict. In a game that revolves around knowing what your opponent is likely to do at any given time, acting erratically makes things a lot more difficult.

Here's Dendi in The International 3, against the Chinese team Tongfu. You can skip ahead to 47:49 to see the game-changing "fountain hook" by Dendi.



And here's Dendi's match with the bot at the most recent International tournament:



OpenAI has demonstrated not just a solid understanding of the rules of Dota 2, but also an adaptability that's allowed the AI to keep up with players who have made a name stumping other opponents. This may seem like a scary concept—Musk himself is the first to admit the danger of underestimating machines that can out-think people—but there's also a lot of exciting prospects to consider here. Soon, AI bots will be able to provide even more helpful assistance to human decision-making. We already know that bots can be used to find trends in vacation booking that humans hadn't managed to spot, and this is just the beginning. If used properly, this technology can be applied to help all of us live more efficient, balanced lives.

We can treat Dendi's defeat at the hands of OpenAI as another sign of the impending Singularity if we must, but we're far better off looking at this as an opportunity to use advancing AI technology to help improve our lives, rather than assuming that the End is Nigh. At least, until AI technology starts getting better than humans at Dungeons and Dragons. Then we might have a problem.
Science
Science News
Technology
Artificial Intelligence
No