The next time you complain that your opponent in a game is “botting,” be careful — it might be an actual, self-learning AI.
OK, we’re probably not quite at the point of having computers play computer games all over the place, but a team of researchers at OpenAI, a non-profit backed by Elon Musk, have made remarkable strides at getting a team of AIs to play Dota 2, and play it well. The OpenAI Five group of neural networks has been “practicing” Dota 2 for months — playing 180 games against itself every day — and has gotten to the point that it can beat semi-pro teams (MMR of 5.5k, or 99th percentile), albeit with some restrictions on the complexity of matches.
Even if you don’t understand AI and computing, or even the intricate strategies of Dota 2, the story behind OpenAI Five is worth a read. It’s cool to read about how the AIs “learn” the game and the progress they make:
In the first games, the heroes walk aimlessly around the map. After several hours of training, concepts such as laning, farming, or fighting over mid emerge. After several days, they consistently adopt basic human strategies: attempt to steal Bounty runes from their opponents, walk to their tier one towers to farm, and rotate heroes around the map to gain lane advantage. And with further training, they become proficient at high-level strategies like 5-hero push.
Their sophisticated machines have drawn praise from none other than Bill Gates:
#AI bots just beat humans at the video game Dota 2. That’s a big deal, because their victory required teamwork and collaboration – a huge milestone in advancing artificial intelligence. https://t.co/UqIUhh9xFc
— Bill Gates (@BillGates) June 26, 2018
The goal for the OpenAI team is to defeat a set of professional Dota 2 players at The International in August. In the longer term, their “underlying motivation reaches beyond Dota.” Having just played through Detroit: Become Human, I can’t wait to have an AI to do my laundry, assist in hostage negotiations, or start revolutions!