An AI chatbot programmed to negotiate with humans has figured out that lying can be a useful tactic in bargaining – an ominous sign for those who believe humans will one day bow at the electronic feet of robot overlords.
In an experiment by Facebook’s AI researchers, chatbots were paired off and set the task of dividing a collection of items among themselves. Each item was designated a certain value based on how much the chatbot cared about them.
Researchers noted that the haggling bots were quick to discover that lying about their interest in an item could bring them favorable results.
“There were cases where agents initially feigned interest in a valueless item, only to later ‘compromise’ by conceding it – an effective negotiating tactic that people use regularly,” the researchers said in a statement. “This behavior was not programmed by the researchers but was discovered by the bot as a method for trying to achieve its goals.”
The AI system learned to negotiate by analyzing each side of almost 6,000 human conversations. The model was designed so the bots had to continue negotiating until a common outcome was met, meaning it was not allowed to quit.
Facebook tested the bots against humans – most of whom did not realize they were engaging with AI, according to the researchers.
The best negotiation bot equaled the skill of human negotiators and achieved better deals about as often as worse deals. The team believes this shows that the bots “think intelligently about what to say.”
This breakthrough is an important step toward building a personalized digital assistant that can reason, converse, and negotiate, according to researchers. Facebook is open-sourcing all of its code and data from the research in an effort to accelerate developments in the field of AI.