Technology is always advancing in today’s world and many feel that the human race will always be ahead of it, but there is the possibility that it may even overtake us. There have been numerous movies about computer systems being able to outsmart their human creators and The Terminator franchise is probably the most well-known, but what if it was real? That is what Facebook had to deal with after having to shut down an artificial intelligence (AI) after it developed its own language.
In a bit of an interesting, and somewhat frightening, story, Facebook recently had to shut down an artificial intelligence system they were working on because it got out of hand. It isn’t that the work start to become too much or that it was too hard, but Digital Journal revealed that the AI actually stopped using English and began using its own language that it created.
While it isn’t the first time this has happened, it was concerning enough for the developers that they needed to cease working on it.
According to Fast Co. Design, the new AI from Facebook created a language that may have come across as nonsense and gibberish to humans, but the computer understood it. Not only was it able to understand its own new language, but it used it to communicate with other artificial intelligence systems to complete the tasks assigned to them.
In a primary example from the Facebook development, there were two artificial intelligence systems which were named Bob and Alice. When given a task, the two systems began communicating with one another, but in terminology that doesn’t make a whole lot of sense to people.
Bob: “I can can I I everything else.”
Alice: “Balls have zero to me to me to me to me to me to me to me to me to.”
All of that may appear as nonsense and quite repetitive but Bob and Alice were able to understand what needed to be done. Dhruv Batra, a research scientist from Georgia Tech who was visiting Facebook AI Research (FAIR) and he said the AI agents will “drift off understandable language and invent codewords for themselves.”
People have seen this type of thing happen in Hollywood with The Terminator franchise, Resident Evil, I Robot, and many others. When The Terminator originally came out in 1984, it seemed as if that kind of technology was so far away from ever becoming a reality. Here it is, more than three decades later and it isn’t only a reality, but it is even stronger and more advanced.
The problem at hand may not be that artificial intelligence is developing its own language, but that humans aren’t attempting to learn them. People teach and design AIs to learn, talk, and communicate, but there is no learning coming from the human side in hopes of opening more lines of conversation.
As Fast Co. Design states, though, the bigger problem is that allowing AIs to develop their own language could prove to be catastrophic. Humans may never have any idea what the agents are saying in full and be left in the dark about what is truly going on.
Developers at Facebook had to work on getting the AI back on track by putting the communication language back into English. The only problem is that the agents know enough to go out on their own as needed, and that could cause numerous projects to be shut down time and time again. The world of The Terminator franchise is one that always seemed like a futuristic world that was too far off to exist in our lifetime.
The big difference is that it has come and it is here now.
[Featured Image by Orion Pictures]