DeepMind: AIs have the potential to become ‘aggressive’ or work in teams

Author

Sam Shead

February 15, 2017

DeepMind ANTS

DeepMind

Artificial intelligence (AI) agents have the potential to become aggressive or work in teams, according to researchers at DeepMind.

A paper released by five computer scientists from the London-based company, which is owned by Google, used games to look at how AIs behave alongside one another.

Joel Leibo, a research scientist at DeepMind and the lead author on the paper, told Business Insider on Thursday: “We were interested in the factors affecting cooperation.”

When asked about AI aggression, Leibo stressed: “We have to be careful not to anthropomorphise too much. These are toy problems aimed at exploring cooperative versus competitive dynamics.”

Describing the study in a blog post on the DeepMind website, the researchers said that they used two basic video games called “Wolfpack” and “Gathering” to analyse the behaviour of AI agents.

“We needed two different research environments in which we could vary aspects of the learning challenge between them and observe their impact on the emergence of cooperation,” Leibo told Business Insider. “For example, in Gathering it is easier to learn to implement a cooperative policy while in Wolfpack, it is harder. This difference cannot be captured by the classical models of social dilemmas in game theory.”

Gathering is a two-player game that requires each player to collect as many apples as possible from a central pile. In the game, players have the option to “tag” their opponent with a laser beam so they can no longer collect apples.

The second game, “Wolfpack,” requires two players to find a third player in an environment littered with obstacles. It’s possible to earn points either by capturing the prey or by being close to the prey when it is caught.

The DeepMind researchers found that AIs can behave in an “aggressive manner” or in a team when there is something to be gained. They also noted that AIs altered their behaviour to become more friendly or antagonistic depending on the situation in the game and what was at stake.

In the DeepMind blog post, the researchers wrote: “As a consequence [of this research], we may be able to better understand and control complex multi-agent systems such as the economy, traffic systems, or the ecological health of our planet — all of which depend on our continued cooperation.”

 

 

This article was written by Sam Shead from Business Insider and was legally licensed through the NewsCred publisher network.

Great ! Thanks for your subscription !

You will soon receive the first Content Loop Newsletter