• Choose a category
• All
• Classic-Fiction

By Root 1684 0

toppling or getting stuck. Such a system, in order to drive itself successfully, has to have a nontrivial storehouse of prepackaged knowledge structures that can be selectively triggered by the scene outside. Thus, some knowledge of such abstractions as “road”, “hill”, “gulley”, “mud”, “rock”, “tree”, “sand”, and many others will be needed if the vehicle is going to avoid getting stuck in mud, trapped in a gulley, or wedged between two boulders. The television cameras and the rangefinders (etc.) provide only the simplest initial stages of the vehicle’s “perceptual process”, and the triggering of various knowledge structures of the sort that were just mentioned corresponds to the far end, the symbolic end, of the process.

I slightly hesitated about putting quotation marks around the words “perceptual process” in the previous sentence, but I made an arbitrary choice, figuring that I was damned if I did and damned if I didn’t. That is, if I left them off, I would be implicitly suggesting that what is going on in such a robot vehicle’s processing of its visual input is truly like our own perception, whereas if I put them on, I would be implicitly suggesting that there is some kind of unbridgeable gulf between what “mere machines” can do and what living creatures do. Either choice is too black-and-white a position. Quotation marks, regrettably, don’t come in shades of gray; if they did, I would have used some intermediate shade to suggest a more nuanced position.

The self-navigation of today’s robot vehicles, though very impressive, is still a far cry from the level of mammalian perception, and yet I think it is fair to say that such a vehicle’s “perception” (sorry for the unshaded quotation marks!) of its environment is just as sophisticated as a mosquito’s “perception” (there — I hope to have somewhat evened the score), and perhaps considerably more so. (A beautiful treatment of this concept of robot vehicles and what different levels of “perception” will buy them is given by Valentino Braitenberg in his book Vehicles.)

Without going into more detail, let me simply say that it makes perfect sense to discuss living animals and self-guiding robots in the same part of this book, for today’s technological achievements are bringing us ever closer to understanding what goes on in living systems that survive in complex environments. Such successes give the lie to the tired dogma endlessly repeated by John Searle that computers are forever doomed to mere “simulation” of the processes of life. If an automaton can drive itself a distance of two hundred miles across a tremendously forbidding desert terrain, how can this feat be called merely a “simulation”? It is certainly as genuine an act of survival in a hostile environment as that of a mosquito flying about a room and avoiding being swatted.

Pondering Dogthink

Let us return to our climb up the purely biological ladder of perceptual sophistication, rising from viruses to bacteria to mosquitoes to frogs to dogs to people (I’ve skipped a few rungs in there, I know). As we move higher and higher, the repertoire of triggerable symbols of course becomes richer and richer — indeed, what else could “climbing up the ladder” mean? Simply judging from their behavior, no one could doubt that pet dogs develop a respectable repertoire of categories, including such examples as “my paw”, “my tail”, “my food”, “my water”, “my dish”, “indoors”, “outdoors”, “dog door”, “human door”, “open”, “closed”, “hot”, “cold”, “nighttime”, “daytime”, “sidewalk”, “road”, “bush”, “grass”, “leash”, “take a walk”, “the park”, “car”, “car door”, “my big owner”, “my little owner”, “the cat”, “the friendly neighbor dog”, “the mean neighbor dog”, “UPS truck”, “the vet”, “ball”, “eat”, “lick”, “drink”, “play”, “sit”, “sofa”, “climb onto”, “bad behavior”, “punishment”, and on and on. Guide dogs often learn a hundred or more words and respond to highly variegated instances of these concepts in many different contexts, thus demonstrating something of the richness of their internal category systems (i.e., their repertoires