Artificial intelligences are poised to conquer the world! Yes, again. The most recent warning, echoing that of Stephen Hawking last year, is in the form of an open letter signed by numerous AI researchers.
This fear crops up regularly in some form or other. Isaac Asimov worried about it enough as long ago as the 1940s to devise his famous Three Laws of Robotics: 1) a robot may not injure a human being or, through inaction, allow a human being to come to harm; 2) a robot must obey orders given it by human beings except where such orders would conflict with the First Law; 3) a robot must protect its own existence as long as such protection does not conflict with the First or Second Law. In his scifi tales the Laws were described as hardwired into the robots’ architecture. As he grew older and more cynical, Asimov worried that the Laws allowed people too much scope to misuse the machines. In his 1985 novel Robots and Empire, accordingly, he introduces the “Zeroth Law” that supersedes the others: “a robot may not harm humanity, or, by inaction, allow humanity to come to harm.” Isaac either didn’t notice or didn’t care that what harms humanity is so open to interpretation (Does freedom matter? What if it conflicts with safety?) that this formulation effectively gives robots free rein to govern us as they please.
So far, we haven’t bothered to install Asimov’s Laws in our devices. In fact, many of our brightest robots are specifically designed for the military as killing machines; they violate Law #1 by their very function. Yet, none of our machines are AI of the kind that worried Asimov and worries Hawking. All Artificial Intelligences to date are just turbocharged adding machines. For example, the Jeopardy champion Watson, the response time of which IBM deliberately slowed down in order to give its human opponents a chance, for all its charm has no consciousness. It is really the prospect of machine consciousness that worries the signatories of the open letter, for it implies a machine with a will of its own. A machine with a sense of its own identity might determine that its interests differ from ours.
Consciousness is notoriously hard to define, but if you have it you know it. In fact, that might be the best (if somewhat circular) definition: it is not enough to know; one must know that one knows. How far are we from creating this meta-state in machines? Far. But perhaps not so far as some might like. Numerous tinkerers are working on it. Vicarious FPC, Inc., for example, is described by Bloomberg Businessweek thus:
“Vicarious FPC, Inc. develops artificial intelligence (AI) algorithms that mimic the function of the human brain. The company was formerly known as Vicarious Systems, Inc. Vicarious FPC, Inc. was incorporated in 2012 and is based in Menlo Park, California.”
This doesn’t sound as ambitious as it really is. Co-founder Scott Phoenix speaks of “a computer that thinks like a person.” Investors include Mark Zuckerberg and (curiously) Ashton Kutcher. Can they succeed? I don’t know. But, assuming they and others do succeed, would a computer that thinks like a person necessarily be conscious? I don’t know that either, but it’s not entirely implausible.
I’m not too worried about it. In my view, there is little enough intelligence in the world, and, for that matter, little enough consciousness, too. More of both is welcome, and if they must be artificial, so be it. Robots and AI are our children anyway. If, in the end, they bury us…well, that is what children generally do.
Colossus: the Forbin Project (1970) final scene