Watson and AI: Does Mind Matter?

The upcoming televised matches between IBM's Watson and Jeopardy champions Ken Jennings and Brad Rutter seems like a good time to reflect on the history of Artificial Intelligence (AI) as a field.

The initial intent of "AI", as defined by John McCarthy, Marvin Minsky, Nathan Rochester and Claude Shannon in their proposal for the seminal 1956 Dartmouth conference was as follows:

The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves.

This describes a project to simulate human intelligence, indeed that project is embedded in the very name of the field. Yet this is not at all what Watson does. It makes no effort to simulate the psychological or neurological processes by which humans answer questions (er, rather, question answers, in the case of Jeopardy). Rather than invoking an associational memory, it instead mines a huge, indexed database of text in search of answers. (This 'inhuman" sort of AI was also found in that other famous electronic John Henry, Deep Blue, which relied on similar "brute force" search and recall capacities well beyond those of any human.)

Does this change reflect a change in the AI program, towards a "pragmatist," problem-oriented intellectual project? Certainly that has been the promotional angle from IBM; this isn't some useless skill like chess, they say, Watson-esque machines will be able to answer real questions and serve people's practical needs.

Or perhaps it is instead reflective of a split in the community - between psychology/philosophy/cognitive science type investigations of mind, and pragmatist/robotics/automation investigations of "action in the world." Musing further still, one could read this as a revival of behaviorist (vs. cognitive) psychology.

Of course this is all quite speculative, it would be great to see more study of these questions. I know of very little work on the history of AI beyond Daniel Crevier's eponymous study and Pamela McCorduck's Machines Who Think--but that may only reflect my ignorance!