Tuesday, June 23, 2015

Eyes of the Machine: He's Back ... to advise us about AI?

He’s back!

A new Terminator sequel comes out next month, and I plan to see it.  Watching the Terminator flicks is a guilty pleasure of mine, which is why I consider this franchise one of the greatest film series in science fiction.  (One caveat: I refer to Terminator I, II, and maybe III, but not Salvation, a movie we’ll pretend never happened.)  Why a guilty pleasure?  Well, as my wife reminded me after I showed her all three movies, they’re basically the same story: killer robots go back in time, wreck havoc, and … that’s basically it.

Even if I grant that criticism (and I do, since my wife may be reading), I’d still say the Terminator character is one of the more complex and fascinating cyborgs in sci fi: sometimes he’s a killer, other times a protector, often a mercurial mix of the two.  You can also identify with the Terminator more easily than other cyborgs, because the movies let you see the world from his first-person subjective viewpoint.  When you see the world from the ‘eyes’ of the machine, you also see that this machine is more than a robot; he’s a cybernetic organism with living flesh and an intelligent mind.  (See the 9 GreatestSci Fi posts for an elaboration on the greatness of the Terminator flicks.)

Herein lies an insight in cinematic art that can carry over from science fiction to science fact: the Terminator is not just Artificial Intelligence (AI) but also Artificial Life (ALife).  Why is that an insight?  Let's briefly consider the story of AI, before and after we got the Terminator…

Before the Terminator:

During the 1950s, 60s, and 70s, research universities like MIT tried to create AI with machines that manipulated logical symbols (sometimes referred to as 'Symbolic AI' or 'Good Old Fashioned AI'—'GOFAI').  These machines gave us lots of useful algorithms, and we can certainly call them ‘intelligent’ in a loose sense.  (See my previous post on how Artificial Intelligence is ‘intelligent’ in this loose senseusually known as ‘Weak AI’).  

However, intelligence more literally understood is a quality of a living mind, and algorithms are not minds per se.  A living, conscious mind, after all, is not something that merely runs arbitrary symbols.  Minds are capable of self-organizing (what system theorists call autopoietic networks), creating meaning (what cognitive scientists call sense-making), and feeling emotions (what neuroscientists call somatic markers).  (If we had Artificial Intelligence with an actual mind that could do these things, it would be known as 'Strong AI.')

After the Terminator:

(Image from Amazon)

Since the 80s—when the Terminator franchise began—AI research in robotics and cybernetics became more pragmatic about what a mind really is: a perceptual (neural) network organized by an active body with feedback loops into an environment.  In sum, mind = brain + body + world.  So if we want AI that entails a mind, then we need embodied cognition, which studies how a mind depends on feedback from a body and an environment.  For example, we have sensorimotor skills that allow us to perceive relevant input, and we feel emotional valences that let us weigh decisions and apply skilled output and control—in fact, think of 'input' and 'output' as demarcations in what philosopher John Dewey called an organic circuit of brain-body-world feedback.

(For the science geeks out there interested in this research, I’d highly recommend the book The Embodied Mind, a major classic in cognitive science; for less technical readers, check out this excellent summary by Scientific American, or read a quick introduction to the topic and its connection to AI on this blog.)

The earlier versions of Symbolic AI or GOFAI fell short of creating actual minds, but cutting research in robotics and cybernetics (e.g., new cybernetics) may come closer to that goal by incorporating data from embodied cognition.  And here we come back to an insight that the Terminator movies reveal:
a living, conscious, intelligent mind is something that literally needs to be fleshed out.