Following The International championship on Saturday, the system was first to beat a real player at a comeptition
In 1996, IBM computer Deep Blue beat chess world champion, Garry Kasparov. It was the first computer to beat a human opponent at a sports or gaming competition. Fast forward 21 years to this weekend's The International, and the Elon Musk backed OpenAI defeated Danil "Dendi" Ishutin at Dota 2 in a competitive 1v1 game. You can see the match below at the 6:30 mark of the video.
Elon Musk's bot, OpenAI was trained for the match by playing against itself and even came out to full fanfare and dressing gown, which is probably a vision on Douglas Adams could have imagined. Bot games are of course common and a 1v1 game of Dota 2 is nowhere near as involved as the traditional team based 5v5 game, but the OpenAI bot came out victorious.
“So we didn’t hard-code in any strategy, we didn’t have it learn from human experts, just from the very beginning, it just keeps playing against a copy of itself. It starts from complete randomness and then it makes very small improvements, and eventually, it’s just pro level,” said researcher for OpenAI Jakub Pachoki.
The bot has played and failed so many games thanks to its computational power that it has effectively outplayed the potential strategy of a human player. The training of the bots takes about an hour to be trained enough to beat the in-game bots and after two weeks was at the level displayed in the match OpenAI also admitted they are working on a 5v5 version of the bot. Elon Musk tweeted about the victory later that day.
OpenAI first ever to defeat world's best players in competitive eSports. Vastly more complex than traditional board games like chess & Go.— Elon Musk (@elonmusk) August 12, 2017
Musk was also appreciative of the help and power of cloud computing in order to make this a reality.
Would like to express our appreciation to Microsoft for use of their Azure cloud computing platform. This required massive processing power.— Elon Musk (@elonmusk) August 12, 2017