Aleatorist has been kind enough to draw attention to DARPA’s Big Dog robot in response to my brief post about the E.A.T.E.R Robot.

It just so happens that I watched a documentary on this very robot and others yesterday. Apparently the advantage of legged robots such as this is that they can go just about anywhere. Where robots that use wheels and treads get bogged down in mud and rocky terrain, legged robots are able to tackle just about any terrain. Just think about the billy goat, for example. Even billy goats that are a few days old are able to climb mountainous terrain with little trouble. The remarkable thing about Big Dog is just how biological its movements look. You get a sense of this when you watch it slipping on the ice in the film above, but the footage I saw yesterday was even more remarkable. There researchers kicked the robot from the side as it was walking and the manner in which it recovered was eerily organic. If the Turing Test were shifted from being a test for intelligence to a test for life, there would certainly be a sense in which it would be difficult to distinguish Big Dog from a living organism.

In a related vein, the learning robots are truly remarkable as well. This video talks about the basic principles behind the “programming” of these robots:

Heavily influenced by research in neurology, these robots do not begin with ordinary pre-defined programs with sets of rules for executing a set of activities, but rather begin with neural networks that evolve executable programs through trial and error imitation through which the robot comes to learn how to do certain actions. In other words, the system learns. The research in neurology indicating neural plasticity, coupled with these sorts of AI projects, is one of the major reasons I often find myself so reactive to models of mind such as we find in Kant’s first Critique, Fodor’s functionalism, or Chomsky’s deep grammar. In each of these cases we begin with a pre-defined and rigid “program” that simply subsumes input under a set of pre-existing categories, functions, and rules. The programming is not something that itself really develops or emerges. It seems to me that a revolution would take place in philosophical epistemology if we were to shift the questions of epistemology away from knowing to questions of learning, investigating the manner in which systems move from fuzzy and disorganized states to patterned and organized states. It is to his credit that Deleuze saw learning as being of greater significance than knowing in Difference and Repetition. Dewey was also similarly sensitive to the significance of learning over knowing throughout his work as well.

Advertisements