Sunday, July 15, 2012

Art from a Machine

In a recent episode of Radiolab, Jad and Robert sampled some of the more intriguing views of Alan Turing.  One of Turing's most exciting insights was that we humans are essentially just biological machines.  From there, they entertained a brief discussion on whether or not a machine could ever produce a work of art.

To create a work of art, a machine would need to have an understanding of concinnity.  That is the perfect union of all parts of a work of art: its form, its function, its ability to inspire new thoughts, its originality (novelty), etc.  Concinnity is the harmonious involution of all those things that make a piece profoundly beautiful.  But a machine could pass a Turing test and still not be capable of recognizing beauty on any level.  Nor could it necessarily be capable of empathy.

"The highest accomplishment of the human consciousness is the imagination, and the highest accomplishment of the imagination is empathy."  - Dean Young (The Art of Recklessness)

I do agree that we are biological machines, albeit marvelously intricate and wildly capable ones, the likes of which have never been seen before on this or any other known world.  However, simply because a machine can be mistaken for a human in casual conversation does not mean that it is capable of anything more than that very thing.  In fact, Turing concedes as much in his paper "Computing Machinary and Intelligence", which is where he first proposed the test.  The point of the Turing test is to prove that a machine is capable of thought, for that definition of thought given in the paper's introduction.

Now we are at a point in time when we are seeing machines capable of testing that barrier more earnestly each year.  It will soon be broken through, but we will have still only barely scratched the surface of the field of Artificial Intelligence.  I think that when the test is finally passed for the first time, there will be a frenzy of sensationalistic media coverage about "machines that can think," "human exceptionalism," etc.  But that will be a classic misinterpretation of scientific results.  The question that most people who are not AI researchers will be thinking is can we make a machine that behaves like a human being, not can we make a machine that is capable of passing the Turing test.

For me, and I'm sure many, the true test of a machine's ability to behave like a human will be to feel empathy and interpret beauty.  It may be possible that one is a necessary condition for the other.  Those are barriers that I do not think we will live to see breached, although I would be excited to be proven wrong.

Put in that perspective, it seems like being able to build a machine that could hold a convincing conversation should be mere child's play.  A vastly more impressive accomplishment will be building a machine that is capable of works of art, and even more that is capable of building other machines that are capable of works of art, ad infinitum.  Of course, even then it will have been a HUMAN accomplishment.

No comments: