CEO of Tesla Motors, SpaceX, and supporter of Open AI, Elon Musk, has many concerns about artificial intelligence. Perhaps it is surprising to find that Musk believes that the government should regulate AI before it is fully developed. He thinks that if the government is reactive, as it usually is, they will be too late to prevent whatever cataclysmic disasters might come our way.
Musk differentiates between narrow AI and deep AI. Narrow AI is the kind of learning machine meant to perform a specific task. For example, he considers a self-driving car narrow. General learning machines, he says, are a bigger threat. These deep AI machines could evolve beyond anything we can comprehend.
While Musk doesn’t believe that machines will necessarily become evil, he does see the possibility. And perhaps more importantly, he seems to think that if AI is controlled solely by a small group of people, then there is the possibility for those people to abuse its power. Either scenario is quite frightening.
Facebook’s Mark Zuckerberg has openly opposed Musk on AI. Zuckerberg says that he doesn’t understand the doomsday warning and commented, “It’s really negative and in some ways I actually think it is pretty irresponsible.” Musk took to Twitter, to reply, “I’ve talked to Mark about this. His understanding of the subject is limited.
Quite recently, Facebook pulled the plug on an AI project that at first glance seems pretty meaningless. The AIs dubbed Bob and Alice were supposed to be learning how to trade balls, hats and books. The goal of the project was to negotiate by mimicking human trading. Instead, the AIs developed their own trading language. It looks like garbage:
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 me to me
Bob: you i everything else . . . . . . . . . . . .
Alice: balls have a ball to me to me to me to me to me to me to me
Bob: i i can i i i everything else . . . . . . . . . . . .
Alice: balls have a ball to me to me to me to me to me to
Bob: i . . . . . . . . . . . . . . . . . . .
Programmers say that machines developing their own forms of communication should be expected. Some people warn that this is terrifying news, while others say that it is just how machines interact. However, if experts are unable to decipher these messages between computers, is that not of some concern?
In conclusion, I don’t think that Musk is being alarmist. He is directly involved in developing AI. He isn’t asking for everything to be shut down, he just wants to see checks and balances in place. With the creation of something that will soon be smarter than humans, that sounds like a reasonable request.