UPDATE (June 29, 2016):
The results are in as June 27 and the machines won (at least in my case). NPR recently covered the Dartmouth Digital Arts Exhibition where computers and artists were contested to see if we could judge if it was the work of man or machine. I’ve taken the test and I thought most of them were machines. I’m not sure if that’s me wanting robot friends but needless to say the sonnets were hard to distinguish the first few reads. Try and see if you can do the same!
Machines might be able to produce creative works of art and we are letting them. Not to worry, this isn’t Isaac Asimov’s I, Robot coming into fruition. Coincidentally enough, the three laws of robotics happened around the same time as the Turing Test in 1950. Introduced by computer scientist Alan Turing, the test measures a computer’s capability to perform automatic computing or self-management. The traditional outcome of the test proved that mathematics would not be able to choose or discern computations’ purposes. It shows but does not tell any meaning from crunched numbers and information.
The Dartmouth College Neukom Institute for Computational Science wants to prove otherwise. The goal is to program an artificial intelligence with human intelligence to show a reflection of the self. It might even show where improvements can be made. The Turing Test will disguise the computer as human to see if the computer can be as accurate as humans without human input. Three tests in the arts will be administered and judged: DigiLit, PoetiX, and AlgoRhythms, each attempting to match the human quality of a short story, sonnet, and dance music mix respectively.
Again, the study is not out to replace humankind. The tests will see how comparable a machine is to a person when creativity is concerned and if a computer can trick us into thinking a human’s output of work was given. AlgoRhythms seems to be the forerunner of the tests as far as computing goes. I don’t imagine much technique involved or taking too long to compute desirable frequencies and sounds. If the robotic shoe fits… As for DigiLit and PoetiX, there is more to be said. Words are ambiguous and deal in multiplicity. The preferred word or choice word might have a favored definition over another. Just the same, words might be synonymous with other words that could serve as replacement and that may be where the machine falters. Sonnets would have a better chance of concealing identity given the rhythm and line count of iambic pentameter. Short stories are more predictable as a single word or phrase could reveal the storyteller.
Is that a reflection of the self then? If we give the machine human input to start computing on its own, it is arguable to say that the machine was never truly anonymous like a human to begin with. The computer is built by a human and is therefore only made durable as the human who made it. I wouldn’t say this is a loophole or obvious limitation on the computer’s behalf. Sentient machines have been talked about and it sounds like they can only be made sentient if they have that point of reference, that initial human input. One machine would have the quality of being human but not become the quintessence of human beings. It indirectly creates these other sentient machines but from that first human’s input, forming a recycled pattern of human similarities, not actualities. The potential conflict is man against man, but is vicariously lived through machine against machine. Then there’s the uncanny valley, but that’s a different kind of unnerving.
So does this tin-man have a heart?
Submissions were due April 15 and from The Washinton Post‘s Nancy Szokan, the results will be held at Dartmouth’s Digital Arts Exhibition on May 18.