Facebook shuts down AI chat robot ‘after it invents its own language’

Hollywood science fiction is full of depictions of rogue artificial intelligences who go completely off the rails – and a Facebook chatbot just did it for real.

Well, sort of.

Researchers at Facebook’s AI research lab (FAIR) found that two AI chatbots in an experiment had taught themselves their own language without any human input.

The researchers were teaching chatbots to make deals with each other, using human language – but discovered that the bots were communicating with each other in their own, non-human slang.

‘Our interest was having bots who could talk to people,’ said Mike Lewis of Facebook’s FAIR team in an interview with Fast Co Design.

MOST POPULAR STORIES FROM YAHOO UK:

Benefits cheat who claimed £81,000 as she could ‘barely walk’ caught snorkeling in Maldives
Britain’s worst roundabout had SIX crashes just hours after opening
Churchwarden faces jail for having sex with ‘obscene’ life-size child sex doll
One in eight young people admit they’ve never seen a cow
Free movement of EU citizens to Britain will end in 2019

‘Agents will drift off understandable language and invent codewords for themselves.

‘Like if I say ‘the’ five times, you interpret that to mean I want five copies of this item. This isn’t so different from the way communities of humans create shorthand.’

The researchers had to alter the ‘rules’ by which the robots were allowed to communicate.

A researcher said: ‘During reinforcement learning, the agent attempts to improve its parameters from conversations with another agent.

‘While the other agent could be a human, Fair used a fixed supervised model that was trained to imitate humans.

‘The second model is fixed, because the researchers found that updating the parameters of both agents led to divergence from human language as the agents developed their own language for negotiating.’