Pro gamers defeat Elon Musk’s AI Bots in Dota 2

Dota 2

At the International Dota 2 Championships in Vancouver, Canada, a group of artificial intelligence bots played against human team paiN and lost, showing that human players – at least at professional level – still have the edge over computer players.

OpenAI is a research institute that was cofounded by Tesla CEO Elon Musk, and it aims to create human-level artificial intelligence.

At a warm up match earlier in the month, the bots managed to defeat a human team. However, in the latest game OpenAI’s bots managed more kills than paiN, but they were unable to apply strategies and squandered opportunities, giving the human team a win.

As Wired reports, Mike Cook, a games an AI researcher at the University of Falmouth, UK, tweeted what he thought were the flaws in the bot's game.

Using games to test AI

Pitting artificial intelligence against human players in sports and games has a long history. Deep Blue, a chess-playing computer developed by IBM, was the first computer chess-playing system to win a chess game and match against a human world champion in 1996.

OpenAI chose the popular competitive Dota 2 game because it is mathematically more complex than chess, and there are various rules and strategies that players must master to win games.

While technical playing wasn't a problem for the AI’s machine learning, when it comes to producing and changing strategies on the fly, the AI faltered – especially when it came to long term plans.

According to Susan Zhang, a software engineer who worked on OpenAI’s Dota project, “They simply don't have any mechanism to ‘plan’ for more than 14 minutes at a time…This definitely contributes to the lack of long term strategy.”

The researchers behind OpenAI aren’t too despondent over the recent loss, with the team telling Wired that it was confident its bots would beat a pro team soon.

The International Dota 2 Championships offer one of the largest cash prizes in epsorts, with a huge $25 million (£20 million, AU$35 million) up for grabs.