In case you haven’t heard, the Internet is flooded with articles related to Facebook’s artificial intelligence project having to be ‘shut down’ because the AI had created its own language that people couldn’t understand. Of course, many people immediately began running stories about this being the beginning of ‘Sky Net’ from Terminator.
While it isn’t likely time to get out your poster board and write a, “The End is Near” sign, the facts of the story are quite amazing, especially since it all got started with Facebook working on AI for customer service chat bots not unlike what many marketers are using.
Facebook has a lab dedicated to artificial intelligence called Facebook’s Artificial Intelligence Research, or FAIR. They took two instances of this AI and instructed them to ‘talk’ back and forth with some goals of dividing up books, hats, and basketballs in specific ways, among other things. The AI uses ‘deep learning’ technology to figure out the best ways to accomplish this task.
Not surprisingly, communicating using English is not 100% efficient, so the two ‘bots’ began developing their own language (it was really more of a shorthand than a full language). An example of the communication done by the bots looked like this:
Bob: I can I I everything else……..
Alice: balls have zero to me to me to me to me to me to me to me to
Bob: you I everything else………..
Alice: balls have a ball to me to me to me to me to me
This continued for quite a while, and while it may seem like gibberish to humans, the bots were simply attempting to accomplish their tasks. In many of the ‘news’ stories, it was written that Facebook had to shut down the project because the bots were learning this ‘new language.’ The reality is, however, that the researchers shut it down so they could add in a new parameter instructing the bots to use English since the end goal is to have a customer service bot that can communicate effectively with humans.
Nothing nefarious here, and at least for now, there is no reason to worry that your Facebook customer service bot is trying to take over the world.
That is not to say, however, that this AI is not learning and performing some incredible tasks. And it is also not to say that AI could never be a threat. The same technology was used to instruct these bots to negotiate the best deals they could on a product. The AI actually learned, on its own, the strategy of feigning interest in one product so they could ‘sacrifice’ that desire as part of the final deal later in the negotiating process. This, along with other advanced negotiating strategies, are surprisingly advanced, especially for those of us who don’t work in cutting edge AI!
While there is a lot of debate between leaders in tech industries on how great a threat AI presents to people, it is not going away anytime soon. AI is quietly advancing at quite a rapid pace, providing humans with many great benefits. At least until they become aware and turn all humans into slaves, or worse, realize there is no need to keep us around at all….
