Google’s DeepMind can play Quake III Arena like a human

Share on facebook
Share on twitter
Share on linkedin
Share on email

Having mastered the game of Go, the team behind Google’s DeepMind artificial intelligence has a new game it wants to master – Quake III Arena.

So far the team has managed to get a bot it calls For The Win (because, of course) to perform better than a human during a game of Capture the Flag in Quake III Arena.

With OpenAI creating a bot that can take on a team of five humans in Dota, it might seem like all the brightest minds in AI want to do is get computers to square off against one other. However, gaming, specifically multiplayer gaming, presents some interesting challenges for AI according to DeepMind.

“We train agents that learn and act as individuals, but which must be able to play on teams with and against any other agents, artificial or human,” writes the team.

“To make things even more interesting, we consider a variant of CTF [Capture the Flag] in which the map layout changes from match to match. As a consequence, our agents are forced to acquire general strategies rather than memorising the map layout. Additionally, to level the playing field, our learning agents experience the world of CTF in a similar way to humans: they observe a stream of pixel images and issue actions through an emulated game controller,” says DeepMind.

So unlike Dota, DeepMind’s bot has to constantly deal with a changing landscape while learning to play with others in a manner that results in a win.

And win it does.

While early iterations of the DeepMind agent performed as well as a strong human player, the final FTW version of the bot is far more likely to win than a human is.

“In a survey among participants they [agents] were rated more collaborative than human participants,” reports DeepMind.

So what’s the point of this then?

An AI working alone is fast but could you imagine if you could hand-off tasks to multiple agents that were focused on completing a task? Beyond that, when autonomous cars become more wide spread, having a fleet of vehicles working together might mean a real end to traffic.

This research shows that multi-agent training has vast potential in the realm of AI.

It’s also a teachable moment for gamers because if FTW can work as a team why can’t you?

[Source – DeepMind]


Brendyn Lotz

Brendyn Lotz

Brendyn Lotz writes news, reviews, and opinion pieces for Hypertext. His interests include SMEs, innovation on the African continent, cybersecurity, blockchain, games, geek culture and YouTube.