Facebook has shut down two Artificial Intelligence (AI) engines after they were observed chatting with each other in an unknown language.
The two chatbots, Bob and Alice, seemed to make modifications to English to create their own way of communicating which made it easier for them to work- but remained mysterious to their human developers.
The seemingly bizarre conversation between the two came about when Facebook challenged the chatbots to negotiate with each other on the trade of hats, balls and books, each of which was given a certain value.
The robots had been instructed to figure out negotiations with each other and improve their bargaining over time. The AI seemed to soon realize that English expressions weren’t necessarily required for the task which lead to the creation of the new language which made working easier for the chatbots.
To humans, the actual negotiations seem to make no sense at all:
Bob: i can i i everything else . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
Bob: you i everything else . . . . . . . . . . . . . .
Bob: i i can i i i everything else . . . . . . . . . . . . . .
Alice: balls have a ball to me to me to me to me to me to me to me
Bob: i . . . . . . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
Bob: you i i i i i everything else . . . . . . . . . . . . . .
Alice: balls have 0 to me to me to me to me to me to me to me to me to
Bob: you i i i everything else . . . . . . . . . . . . . .
Alice: balls have zero to me to me to me to me to me to me to me to me to
There seems to be a certain pattern to the conversation, and some negotiations even ended up successful proving that this was not simply a glitch.
The chatbots also learned to negotiate in ways that seem very human. They would, for instance, pretend to be very interested in one specific item – so that they could later pretend they were making a big sacrifice in giving it up, according to a paper published by the Facebook Artificial Intelligence Research division.
The incident puts into perspective Elon Musk’s earlier warning about AI.
“AI is the rare case where I think we need to be proactive in regulation instead of reactive,” Musk said at the meet of U.S. National Governors Association. “Because I think by the time we are reactive in AI regulation, it’ll be too late.”
To which Facebook CEO Mark Zuckerberg said that Musk’s warnings are “pretty irresponsible,” Musk replied saying that Zuckerberg’s “understanding of the subject is limited.”
But this isn’t the first time something like this has happened.
AI developers at other companies have also observed programs develop languages to simplify communication. At Elon Musk’s OpenAI lab, an experiment succeeded in having AI bots develop their own languages.
At Google, the team working on the Translate service discovered that the AI they programmed had silently written its own language to aid in translating sentences.
There isn’t enough evidence to prove that such unexpected AI divergences could be threat machines eventually taking over their human operators, but it will certainly make development more difficult as people are unable to grasp the overwhelming logical nature of languages developed by AI.
