Over the weekend, Facebook was forced to reevaluate their Artificial Intelligence (AI) systems after developers noticed that it had overwritten human language, and created a language of its own. The easiest way to relate to this, would be the famed “Skynet” portion of the Terminator. Ironically, Elon Musk, CEO of Tesla, recently put out a statement stating that AI is “the biggest risk we face.”
Facebook developed two AI agents, who were to negotiate and communicate with humans in English. These two agents, “Bob” and “Alice,” started communicating in a language that was not understandable by humans. The developers originally made the agents start in English, but they created one much faster and easier allowing them to communicate more efficiently. With this discovery, the developers instantly shut down and modified “Bob” and “Alice,” making them speak solely in English.
Uneasy Development in Artificial Intelligence
When AI agents are using their own language, humans cannot understand exactly what they’re communicating. This leaves humans vulnerable, and unable to determine the incredible intelligence that they do obtain, which might surpass human intervention completely.
It’s important to keep Artificial Intelligence in check and under human control, otherwise AI poses a threat to potentially turn into something that could replace humans entirely.