jps@cat.cmu.edu (James Salsman) (06/02/89)
A very wise psychologist once wrote a bunch of stuff that implied that the psyche was controlled by hydro-mechanical laws. This was wrong, of course, and led to a batch of pseudo-science called "psychoanalysis." Totally bogus stuff. We are more enlightened in these times. We have a branch of pseudo-science called "neurology" that is a composition of mathematics, biology, and information science. This discipline is practiced by a few double-E's, theoretical computer scientists, and psychologists. If some poker writes a book on this stuff updating the principles of Freud in this new scientific context, they will become rich and famous and a new brand of pseudo-science will be born. Let us consider the "shape" of the brain. That is, what would a "map" of the information flow in the brain look like? I found a book with a rat brain direction (in, ahem, vivid colors) in it the other day. This upsets me because I used to have a pet rat named Zak that was smarter than I was, at the time. Zak died when I was 14... my mother cried, but I had sort of been expecting it; he had had a tumor for the past several months. Lab rats seem to have been bred for cancer hypersensitivity by the medical establishment and the FDA. We are the kings and the rats taste our food. Anyway, the Rats brain seemed to be a layer of grey lobes covering a bunch of gland-like lobes. Every different lobe was connected to the spinal cord and the circulatory systems in a different way. I sketched the interconnection topology as best I could, but it won't be high quality data until I get to a better rat-brain picture book [there were references to more medically technical books, some of which appeared to have something to do with humans.] I have read that in humans, the grey matter lobes act as insulation around a somewhat planar network called the "white matter" that is "crunched up," to form a ball. The white mater is not perfectly planar -- it is arranged in layers. The grey matter has fewer neurons than the white matter, but quite a few more capilaries. I am working on an AI project involving the simulation of neural nets that update their topological aspects in real time. This is based on a reflective term-rewriting system. "Reflective" means that the re-write terms act on themselves. In addition, the terms also define a neural net topology. The way that this is accomplished is based on Category Theory. (Elements of categories have state and are called neurons. The "Arrows," or "Mappings," are interconnection schemes, and can be used to build Boltzman machines or Hebb-Marr hypocampial crossbar interconnects, for instance. The trick is that the Category Diagrams represent both the topology of the net and the re-write rules for other parts of the network.) This system is being prototyped on the Connection Machine architecture. Once the low-level routines are complete, I plan on porting a high-level symbolist AI system like Soar into the framework, and adding a wizzy typographical user interface. Maybe a few years down the road DSP voice-processing if the chips get cheap and >>>eventually<<< video. Convincing video *output* is easy, with a Max Headroom puppet running off the serial line of an Amiga. I'm sure glad that parallel hardware is becoming more economically accessible. If CMU doesn't get some quality hardware soon, we will be left in the academic dust. The universities of today should (imho) use SIMD machines for research, develop MIMD machines for the government (since they're so expensive, more money for research and education!) and let the commercial sectors scale the cool software down to serial machines, so that everyone can have access to high-bandwidth real-time full-duplex media. Mass-production of Videophones by the Year 2000!!! The ANSWER to the Postmodern Condition!!! :James P. Salsman Soar Project; Center for Art and Technology; etc, etc. Carnegie Mellon -- :James P. Salsman (jps@CAT.CMU.EDU)