How Alphabet’s DeepMind used a 1999 video game to teach its AI teamwork

Автор: | 06.07.2018

DeepMind, an Alphabet subsidiary, announced Tuesday its efforts to create an artificial intelligence (AI) that functions with human-like performance. Using Quake III Arena’s Capture the Flag (CTF), a 3D first-person multiplayer video game, DeepMind taught AI how to work with and against humans to win the game.

The rules of CTF haven’t changed much since gym class. Two groups of individuals band together to steal the opponent’s flag from their territory while also protecting their own. Teams can tag opponents that enter their territory, sending them back to their respective home base. In Quake III Arena, the team with the most flags captured after five minutes wins.

SEE: How robots and automation may take on the jobs people don’t want to do (Tech Pro Research)

More about artificial intelligence

«From a multi-agent perspective, CTF requires players to both successfully cooperate with their teammates as well as compete with the opposing team, while remaining robust to any playing style they might encounter,» said DeepMind in a blog post. DeepMind also changed the map layout with each game, making the contextual situation even more complex for AI agents.

DeepMind trained a population of agents with a two-tier optimisation process, which allows the agents to form their own goals, said the post. The agents are never given the rules of the game, so they learn how to play the game by their own intuition, which seems to follow the principles of deep learning.

AI agents were mixed in with 40 human players for a CTF tournament. In a short time, agents greatly surpassed the win rate of their human opponents. In fact, the human players even rated the AI agents as more collaborative teammates than humans, according to DeepMind.

If AI is just as functional—and more collaborative—than humans, then who is to say AI can’t be trained to work better with human employees? The fear of human jobs being replaced with AI is not a new phenomenon, but DeepMind’s technology make it even more of a reality. But, it also opens up the possibility that AI could more easily work alongside human employees, rather than just replacing them.

«In general, we think this work highlights the potential of multi-agent training to advance the development of artificial intelligence: exploiting the natural curriculum provided by multi-agent training, and forcing the development of robust agents that can even team up with humans,» said the post.

The big takeaways for tech leaders:

  • Google’s DeepMind used Quake III Arena’s Capture the Flag to create AI that works at a human level, according to DeepMind’s Tuesday blog post.
  • Through the video game, AI learned how to team up with and against humans—as well as other AI—and were even rated as more collaborative than their human counterparts.

Also see


Image: iStockphoto/PhonlamaiPhoto


Добавить комментарий