Vehicle 6: Selection, the Impersonal Engineer

The innovation that Braitenberg explores now is "giving chance a chance" -- he introduces random perturbations in vehicle replication, coupled with natural selection.

His view of the vehicle's world is now a table, where vehicles die -- are selected out of the environment -- by falling off. The designer reaches to the table, picks out a vehicle, and uses a set of nearby supplies to copy the selected machine. The machine and its new copy are then both returned to the table.

Random variation is introduced by requiring the replication process to be fast, which results in sloppy workmanship. Mistakes are made in the copying process; sometimes these mistakes improve the survivability of the vehicle. "This can easily happen when we pick up one vehicle as a model for one part of the brain and then by mistake pick up another vehicle as a model for another part of the brain. Such errors have a much greater chance of transcending the intelligence of the original plan." (NB: Does this imply that the role of natural selection for intelligence is best helped by miscopying of working modules?)

This approach introduces Darwinian evolution to the story, because good mistakes will mean that certain features will be more likely to be present on the table, and will thus be more likely to be passed on to new copies of vehicles.

"Where there has been no conscious engineering at all, as in the case of our type 6 vehicles, analysis will necessarily produce the feeling of a mysterious supernatural hand guiding the creation. We can imagine that in most cases our analysis of brain in type 6 vehicles will fail altogether: the wiring that produces their behavior may be so complicated and involved that we will never be able to isolate a simple scheme. And yet it works."

Mike's End Notes On This Chapter

The notion of intelligence being synthesized along the lines of Darwinian evolution, and the resultant difficulty in providing an explanation of the intelligence that results, is a theme that is becoming more and more common to find in the literature.

For one example, this concept appears in the writings of those who view PDP connectionism as providing a more biologically accurate account of cognition. For instance, in Clark's book "Microcognition", he argues that because the mind is a product of evolution, it is just as likely a place to find a kludge as any. He proposes that we study psychological properties required for success at an evolutionary basic level, because these capacities will impose strong constraints on later solutions to higher-level (cognitive) problems. Clark even details a functional phylogeny of capacities that he believes are good candidates, and which include basic sensorimotor abilities. "Paradoxically, then, the protocognitive capacities we share with lower animals, and not the distinctive human achievements such as chess-playing or story understanding, afford, I believe, the best opportunities for design-oriented insight into human psychology."

A second example comes from research into genetic algorithms, which is currently a sexy topic in some quarters of computer science. In a genetic algorithm, a program for accomplishing something is represented as a string of numerical code. The code is replicated, perhaps crossed with other codes in the environment, and new, mutant programs are produced. The goal of this kind of work is to evolve novel approaches to solving problem. [References here??]

A third example comes from the selectionist camp that is entrenched in some quarters of biological views of cognition. Holders of the selectionist view take as a guiding metaphor the human immune system, which apparently does not create new responses to new infections, but has the response pre-built, waiting to be used. It is selected out by being placed in an environment that requires that immune response to be generated. Following this metaphor, specific perceptual/cognitive responses to the world are pre-existing dispositions, that become realized when they are required. Some straightforward treatments of selectionism by is proponents can be found in Edelman's book "Bright air, brilliant fire" and in Gazzaniga's book "Nature's mind."

Pearl Street | Vehicles Home Page | Dawson Home Page |