Technology Review published this a couple weeks ago: http://www.technologyreview.com/biomedicine/23354/
The author has a good point: just creating a more intelligent machine is not necessarily going to result in it creating an even more intelligent machine. I’ve always said that a funny thing could happen on the way to the Singularity.
More intelligence is not the answer. Computers are already faster at thinking than we are; but they’re not better at it. Of course “better” is relative. In this sense, “better” implies the presence of a wanting to create more intelligent machines, some sort of desire on the part of the AI. Apparently motivation is required to effect the Singularity.
The question is: how do you program wanting, motivation into AI? True motivation. Not just instructions to do something, but instructions to want to do something. We humans seem to have “want” hardwired somewhere in the process of procreation. Apparently our genes give us our motivation. Eating, having sex, being entertained, these are all things that we want to do because the end result is the success of the species. We don’t consciously think: I must eat so I’ll live to the age when I can pass on my genes to my children; I must watch Mad Men so I’ll be in a good mood when my spouse comes home and I’ll be motivated to pass on my genes to my children; or even, I must have sex so I can pass on my genes to my children. We simply have a desire to do these things because our genes code for proteins that build pleasure centers in our brains that are stimulated by satisfying something else in our brains that tell us that we want something. Or something.
We take that desire for granted but it’s tied into the very definition of life. Life is the ability to reproduce oneself (and probably the desire to do it should be included in that definition), thus we do so every chance we get. That’s why we say our desires are hardwired. Being alive makes us want to go forth and multiply even if our day to day decisions don’t seem to have anything to do with procreation.
So what happens when you program this motivation into a robot? If AI wants to procreate as badly as humans and other living things want to procreate, then is that AI alive?
I can’t imagine how to program an AI to actually want something not just because it’s been told to want something but because it truly, in the bowels of its firmware, wants that something. But let’s say it can be done. Can we say then, that this AI is alive in the same way humans are alive?
Is motivation that’s traceable to a subroutine the same as motivation that’s traceable to a gene?
Or conversely: are our wants, our motivations, the very things we cling to as humans, so easily replicated and accounted for that we’re no more than mere robots already? Maybe it’s not a question so much of would the AI be considered alive, but are we as animals actually not alive?
An important question, to be answered as soon as The Singularity hits, which as far as I’m concerned will put us somewhere out in the Twilight Zone.