I was having a conversation with some friends today about the future of Artificial Intelligence and robots. As you probably know, A.I. pervades our lives. Heuristic algorithms run our cars, our homes, and every internet search we do. Mostly these algorithms work behind the scenes, rather than being embodied by an old-fashioned robot with a face. Occasionally something like Apple’s Siri is used as a front-end puppet, so that we can have the entertainment value of “talking with a robot”.
Whether or not this stuff ends up being seen in anthropomorphic embodiments, or manifests mainly as anonymous software lurking in cyberspace, I posited that there is a key question here — one to which we do not yet know the answer.
The question is: will we ever develop A.I. that can solve novel and unexpected problems with something like the facility possessed by humans (and some other species as well). This is precisely what is missing from today’s leading examples of A.I., from Watson to Google Search — the ability to switch contexts, to understand and deal with a radically novel problem, to put together apparently unrelated clues and find a solution within a completely new solution space.
If we do ever manage to forge machines that can cross that chasm of capability, then I think the interesting questions will start. Because at that point our machines might really begin to be our friends or enemies. And then we can even start worrying about SkyNet. 🙂