Facebook AI robots shut down after creating own language humans can’t understand

By @chelean on
Visitors look at the humanoid robot Roboy at the exhibition 'Robots on Tour' in Zurich, March 9, 2013.
Visitors look at the humanoid robot Roboy at the exhibition 'Robots on Tour' in Zurich, March 9, 2013. Reuters/Michael Buholzer

Facebook has reportedly shut down its artificial intelligence robots after they created their own language. Researchers have found that the machines, which are nicknamed Bob and Alice, were communicating with each other in a language that sounded gibberish to humans.

The chatbots appeared to have created their own code words from the English language to make it easier for them to communicate with each other. And while their words appeared nonsense to humans, it is thought that they perfectly understood each other.

Bob: i can i i everything else …

Alice: balls have zero to me to me to me to me…

Bob: you i everything else …

The exchange went on the same way repeatedly. Bob and Alice appeared to be trying to negotiate with each other over a trade. Apparently, the way they were stressing their name was part of how AI robots negotiate. Researchers believe that the chatbots were working out how many of each item they should take. Bob and Alice apparently realised that the scenario did not call for English phrases, and therefore they created a more efficient language instead.

“There was no reward to sticking to English language,” Dhruv Batra, visiting research scientist from Georgia Tech at Facebook AI Research (FAIR), was quoted by Fast Co. Design as saying. “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 shorthands.”

The AI bots creating their own code to speak to each other signal that they are getting smarter. Just like human languages evolve, AI languages also evolve into what they understand better. However, humans would have no clue what the robots were actually saying to one another, and this would be a problem in the future. As sci-fi films like “Terminator,” “The Matrix” and “I, Robot” have warned, AI could take over the world and humans wouldn’t be able to do anything about it.

As Fast Co. Design reports, Facebook ultimately opted to require its bots to speak in plain English. The social media giant admitted that it had no way of understanding any divergent computer language.

“It’s important to remember, there aren’t bilingual speakers of AI and human languages,” Batra said. As people don’t understand how complex AIs think, they can’t see inside their thought process. Thus, an AI-to-AI conversation would only worsen the problem.

Google Translate AI creates own language

Meanwhile over at Google, its AI has also created its own language, and the search engine titan has no problem with it. Last year, Google revealed that the artificial intelligence it uses for Google Translate has created its own codes, and that’s okay.

The AI apparently has become more intelligent that it was trained to be that it seemingly created its own artificial language. The online translation tool started using a neural network to translate between some of its most popular languages. Most of these machine-translation systems only work on a single pair of languages, so there needs to be different systems to translate between unpaired languages. The Google AI, however, can translate between two languages even if it hasn’t been directly trained to do so.

As New Scientist explains, if the neural network has been taught to translate between English and Japanese, and English and Korean, the new system can also translate between Japanese and Korean without going first through English.