Tuesday, June 24, 2014

Less than meets the AI

A mere two weeks ago, the big, breaking news -- apart from the ongoing cascade of geopolitical woes -- was that (in this particular iteration, citing the Washington Post) "A computer just passed the Turing Test in landmark trial."

An under-appreciated genius
Alan Turing is perhaps best known for his role in cracking the German military's Enigma crypto system. Turing thereby -- at the very least -- shortened the war against the Nazis and saved many lives. He also established some of the foundational theorems of computer science. As for the subject/headlines at hand, he speculated, way back in 1950, about artificial intelligence.

Given that the experts struggled -- and still do -- to define intelligence, Turing's insight was (characteristically) brilliant. To wit: don't try to define artificial intelligence; describe its behavior. From which arose the famous Turing Test: if an artificial entity interacting with judges by text messages successfully masquerades as a human at the keyboard, then the entity, too, is intelligent.

Pay no attention to the computer behind the curtain

Such imitation served as the basis of many recent headlines. From that Washington Post article:

... a Russian-made program, which disguised itself as a 13-year-old boy named Eugene Goostman from Odessa, Ukraine, bamboozled 33 percent of human questioners. Eugene was one of five supercomputers who entered the 2014 Turing Test.

Doesn't that 33 percent sound precise and scientific? Not so much. There were three judges, meaning one person was convinced. Fooled. That doesn't seem terribly compelling to me.

The Smithsonian took a deeper, more nuanced look at the situation, in "The Turing Test Measures Something, But It's Not "Intelligence": A computer program mimicked human conversation so well that it was mistaken for a real live human, but "machine intelligence" still has a long way to go."

The 2014 Turing Test ran for only five minutes, while "Eugene" hid behind the clever persona of a teenager for whom English was a second language. From that Smithsonian article:

All of this raises a crucial question: What is it, exactly, that the Turing test is measuring? Some critics have suggested that it is [sic] rewards trickery rather than intelligence. NYU Psychologist Gary Marcus, writing at NewYorker.com, says Eugene succeeds “by executing a series of ‘ploys’ designed to mask the program’s limitations.” Steven Harnad, a psychologist and computer scientist at the University of Quebec in Montreal, was even more skeptical, telling The Guardian that it was “complete nonsense” to claim that Eugene had passed the Turing test. (To his credit, Turing was well aware of this issue; he called his idea “the imitation game,” and spoke of intelligence only sparingly.) Even more awkwardly, the computer, unlike the human, is compelled to deceive. “The Turing Test is really a test of being a successful liar,” Pat Hayes, a computer scientist at the Institute for Human and Machine Cognition in Pensacola, Florida, told me following the 2012 Turing test marathon. “If you had something that really could pass Turing’s imitation game, it would be a very successful ‘human mimic.’”

Beyond those highly cogent objections, I'll propose one of my own: the most essential attribute of what we humans consider intelligence is purposeful volition. However realistically a mechanism, hardware and/or software, responds to stimuli, it's absurd to consider that mechanism to be "intelligent" if, left to itself, it only spins its mental wheels. 

How and why do evolved minds initiate actions? What is free will? What is self-awareness? Alas, no one knows. My suspicion is that -- somehow -- volition is (a) an emergent property of (b) a very large ensemble of quantum states. Within physics as it is presently understood, only quantum mechanics offers any basis for non-determinism.

Alas (as per an earlier article from the Washington Post, "Why quantum mechanics is an “embarrassment” to science"), despite almost a century of effort, physicists still fail to agree what quantum mechanics means. The plurality opinion of a recent survey -- of thirty-three experts -- was: don't ask. The math works.

Image credit: Scientific Instruments
If intelligence does require volition and volition is indeed rooted in quantum indeterminism, I don't find it surprising that intelligence -- whether based in meat or silicon -- remains ill-defined.

Suppose that, someday, our technological toolbox grows to include large-scale quantum computing. (That day may be awhile coming. See, from PhysicsWorld.com, "Is D-Wave's quantum computer actually a quantum computer?") Then, perhaps, an entity will arise with a credible possibility of demonstrating intelligence rather than trickery and mimicry.

Won't that make for an exciting headline? 

No comments: