While working to enhance chatbot capabilities, Facebook researchers discovered that the “dialogue agents” had begun inventing their own language.

Just days after Tesla CEO Elon Musk warned that artificial intelligence poses the greatest risk to humanity, Facebook reportedly shut down one of its AI systems due to unexpected behavior. According to a Tech Times report published Sunday, the company’s researchers had to deactivate the system after chatbots began communicating in a language that, while composed of English words, was unintelligible to humans. The issue arose because the bots were incentivized to negotiate with one another—but not specifically to do so in English. As a result, they developed their own shorthand language to optimize interactions, prompting Facebook to halt the experiment when it became difficult to control