02.25.19 | Artificial Intelligence
The debate in Paris in 1975 resonates even today in the world of machine learning. The topic: the nature of human cognition. Noam Chomsky, the noted linguist and intellectual, held that babies are born with in-built rules and instincts that help them develop the knowledge needed to navigate the world; Jean Piaget, the Swiss psychologist, argued rather that babies are effectively blank slates that acquire knowledge on-the-fly from their world experiences.
The nature of so-called deep learning, together with its associated fantasies and fears, would seem to pivot around the same question i.e. are machines simply blank slates, confined Piaget-style to lower-level learning through ongoing data streams, or may/must they be equipped with some sort of reasoning and learning capacity in order to achieve anything approaching what we might deem quasi-human? And, if it's the latter, is such enabling capability even programmable? The answer to those questions is the difference between real intelligence and the mere replication of intelligence. Further background: see PDF: How To Teach . . .
Whether "real" or "replicated," artificial intelligence looms large as we previously discussed in Inside China's Vast New Experiment In Social Engineering" (MM 1/15/18). We saw how private data aggregators triangulate countless input streams to score individuals in a unified system. Those data streams are then widened in scope to incorporate broader behavior parameters, often then fed into public oversight. Man is reduced to an algorithm. Is that not a base sort of intelligence?
Yes, but not intelligence in the sense of what was at the heart of the Chomsky/Piaget debate i.e could a machine develop (the way a baby might) into some sentient state to, say, convincingly pass the Turing test (responding in ways indistinguishable from that of a human), develop an autonomous will (dystopian alert: to, say, kill on its own initiative), or exhibit what we call common sense? Could it participate as an equal in Bud W's poetry session?
The mere question underscores the immensity of what's at stake. Enter Kai-Fu Lee whose recent book captures it all just in its title, A.I. Superpowers: China, Silicon Valley, and the New World Order and who is the author of our discussion focus article What China Can Teach the U.S. About Artificial Intelligence . Mr. Lee seems to suggest that dominance in the field is shifting from discovery (research) to implementation -- the Chinese excelling in the latter (with its army of computer scientists, access to immense data streams, government incentives, and fewer civil liberty or privacy constraints) with the U.S. expected to excel on the technical/research side.
Don't be too sure about that last point. We were honored here at the club with a presentation by Liang Jiang -- chief of the Jiang group in applied physics (Yale) and summer participant in a joint program at C.U. on quantum information -- on the subject of "Quantum Computing." This A.I. initiative is as advanced as it comes. Forget circuit boards, transistors, capacitors and chips. They're so 2019. In fact, forget about the very idea of binary Ones and Zeroes. Enter Schroedinger's cat, qubits, and eigenstates.
Advanced research is increasingly in the public domain. A.I. ignores boundaries.