The Infamous Facebook AI Language Experiment.
Mark Zuckerberg, the CEO and founder of Facebook and Elon Musk the CEO of Tesla Motors and SpaceX are reported to to have two very different opinions of artificial intelligence. Zuckerberg has been quoted as saying that anyone who focus’s on the danger of AI is doing a dis-service to humanity, while Musk is very outspoken on what he considers to be huge risks associated with it.
While just because these two entrepreneurial billionaires have strong opinions does not be mean either of them has insight into what might happen as a result of AI research. it does bear listening to. Especially since it seems that public opinion on AI is very mixed and that maybe we do have a fear of AI that stems from negative science fiction books and movies.
I say this because recently, Facebook was experimenting with AI with the goal of creating a program that could interact with humans in a better way. For example an AI program could try to sell a person insurance or sell them a car, basically holding a normal two way conversation with real people. In order to speed the process of developing this language capability, Facebook engineers used two AI programs to talk to each other. Seems like a good idea right?
The Myth of Ill Fated Facebook AI Experiment.
The story goes that after some time, these two AI entities started to refine the language used and developed a short hand language that soon became meaningless to their human developers. Facebook shut it down, stating that the purpose was to develop human language skills not a new machine to machine language.
Since then, the rumors and headlines have been supporting wild speculation. Did the machines develop a new secret language and Facebook had to shut it down before it got out of control? Were these AI programs now so advanced that they are aware and secretly plotting their own agenda? Surely Facebook AI researchers must be hiding the real truth…
What I have learned from this is that at lease at the level of the media, there is more interest in the dangers of AI than the benefits. I wonder if we will see more of this type of paranoia as the science progresses?
Recently Hansen Robotics showed of their new AI robot Sophia. She ( it?) made a comment or two that I think was meant to be funny about how AI was plotting to take over the world. Rather than point out the potential of this level of AI development, much of the press has focused on this, trying to spin it as a real threat from one of the first and most public AI representatives.
The truth is that the state of the art for artificial intelligence is such that programs today can do all computational activities such as math, chess or GO, faster and better than humans. AI and robotics together will likely make possible the replacement of millions of humans in jobs ranging from taxi and truck drivers to lawyers, doctors and journalists.
Maybe this is what we should we should be concerned about first?