Facebook has shut down a number of its AI systems after they developed their own language in order to communicate more efficiently. It's no secret that Facebook is investing heavily into AI, with one of its goals being to create AI that can successfully communicate with people using English. It appears that one of the company's latest attempts didn't exactly go according to plan, with a set of chatbots developing their own communications system that humans couldn't understand.
The AI chatbots were initially provided with rewards for communicating with each other. When it came to solely communicating in English, though, no rewards were set out, which led to the chatbots gradually creating their own language. The new language was still based on English, though it was constructed in a manner that was no longer understandable to humans. The reasoning behind this development is relatively straightforward; over time, the chatbots started developing codewords or phrases that only they understood so that their communication could be conducted in a more efficient, quicker manner. However, it appears that the project went too far and Facebook has now shut down the AI system. These weren't the only results of the project; the chatbots also learned to negotiate and even began showing interest in certain items in order to improve the outcome of a larger deal later on, the company said.
Facebook's latest project shows potential in the context of general advancements in machine learning, but all of this could come at a cost. Artificial intelligence is still a relatively new technology, but considering how an AI-powered system can develop a language so early on, the possibilities of what it could eventually be capable of are a worry for some experts in the industry. Many fear that AI could eventually turn against humans and therefore have unforeseen consequences. There are certainly a number of risks when it comes to the technology, though some do argue that all technology has its drawbacks. What remains to be seen for now is how developers manage other similar obstacles in order to safely evolve the technology and whether governments all over the world start regulating AI in a stricter manner.