Facebook Shuts Down their Artificial Intelligence Robots after they Began Conversing in an Unknown Language


The giant social media and technology company, Facebook, was forced to stop an experiment after their artificial intelligence robots started chatting in a language only known to them. The Facebook AI researchers had built the chatbots earlier this year. They were meant to learn how to negotiate with customers by simulating human trading and exchange language.

The two chatbots named Bob and Alice were challenged to negotiate with each other over a trade involving hats, balls, and books. Each of these objects was assigned a particular value. However, the two chatbots broke down since Facebook researchers had failed to assign a reward for conducting the trade in English. The chatbots, therefore, developed their own shorthand terms for the deal and conversed in a language that could not be understood by those manning the system.

Though the conversation between Alice and Bob appears to hold little sense, it seems like it had some rules of speech. The way these two chatbots kept stressing on their names seems not to be a glitch but rather a way of conveying a message to each other. Researchers say that the negotiations appeared to have ended successfully despite having been conducted in a strange language.

When Dhruv Batra, a Facebook AI researcher, was invited to speak on the matter, he said that the bots might have developed a shorthand to facilitate the negotiations. Just like agents use broken speech and adopt codewords during trade, the bots might have just done the same according to Batra. Mark Liberman who is a linguist, however, commented that the language used by the AI couldn’t have been a precursor to new forms of human communication. He argued that people usually would accompany shorthands and codewords with bodily gestures to facilitate speech. The robots’ conversation, however, was just text-based.

One of the researchers, Mike Lewis, attempted to mop out the rumors claiming that Facebook had shut down the chatbots because they panicked, or for fear of the results. He expounded that their interest was to create chatbots that would converse with people, and that hadn’t been the case with Alice and Bob. They terminated the experiment because they expected the bots to behave differently.

This isn’t the first case where AI systems create their language. Earlier this year, the Google Translate AI also created its own language which it could translate into and out of. The company, however, embraced it as a healthy development and allowed it to thrive.


Please enter your comment!
Please enter your name here