AIs which train on large datasets alongside other AIs are increasingly producing behaviours unintelligible to humans.

There is a growing concern that human-machine cooperation will break down as AI becomes more complex and it develops behaviours amongst itself.

In controlled experiments in popular games such as Chess to Poker to Go, AI develops intricate solutions which are incompatible with human social cues and cognition.

PhD Candidate in Artificial Intelligence at Oxford University, Ravi Hammond, explains why human and AI cooperation is one of the biggest challenges we face.

Image: Shutterstock: PopTika

Produced By: Demos Hammond

Featured In Story: Ravi Hammond, PhD Candidate in AI at Oxford University

First aired on The Wire, Wednesday 31 July 2024