lishka@uwslh.UUCP (Brain-fried after too much hacking) (01/03/90)
bwk@mbunix.mitre.org (Kort) writes: >In article <1037@ra.stsci.edu> bsimon@stsci.EDU (Bernie Simon) writes: > > 6) While there are good reasons to believe that thinking is a physical > > activity, there are no good reasons for believing that thinking is the > > execution of a computer program. Nothing revealed either through > > introspection or the examination of the anatomy of the brain leads to > > the conclusion that the brain is operating as a computer. If someone > > claims that it is, the burden of proof is on that person to justify that > > claim. Such proof must be base on analysis of the brain's structure and > > not on logical, mathematical, or philosophical grounds. Since even the > > physical basis of memory is poorly understood at present, any claim that > > the brain is a computer is at best an unproven hypothesis. >The brain is a collection of about 400 anatomically identifiable >neural networks, interconnected by trunk circuits called nerve bundles, >and connected to the outside world by sensory organs (eyes, ears, nose, >tactile sensors) and effectors (muscles, vocal cords). Neural networks >are programmable computational devices, capable of categorizing stimuli >into cases, and capable of instantiating any computable function (some >more easily than others). Artificial neural networks are used today >for classifying applicants for credit or insurance. They have also >been used to read ASCII text and drive a speech synthesizer, thereby >demonstrating one aspect of language processing. As to memory, you >might want to explore recent research on the Hebb's synapse. Although Mr. Kort does not specifically say so, he seems to imply that the brain is a computer because it "is a collection of ... neural networks" and because "neural networks are programmable computational devices...." However, a *very* important distinction is not brought out in his argument: although artificial neural networks resemble human neural networks, they are only *very* crude models of the real ones, if they can be called that at all. Time and time again I see people make this mistake: because artificial neural nets were designed with some very *basic* features that human neural networks have, then the human neural networks must be like the artificial ones. This is wrong! I have a B.S. in computer science (with an emphasis on AI), and I was really gung ho about neural networks during my undergraduate years (I still think they are a very promising area of AI research). Therefore, to broaden my knowledge of neural nets, I dove head-first into a two course neurobiology series, intended for biologists and medical students. My final paper in the second neurobiology course (and also for a graduate level AI course) was on areas in neurobiology that were promising to AI artificial neural net research (the main focus was on some computer simulations of opposum olfactory cortex using artificial neural nets). These are my qualifications, which I admit are minor compared to other peoples'. However, I think that I have the knowledge to compare artificial and real neural nets for my argument. One of the most important things I learned was that many AI researchers are fooling themselves if they think that current artificial neural nets are anywhere close to real live neural networks in living brains. It is almost like comparing apples and oranges. Most artificial neural networks use relatively simple threshold functions (typically only one type per neural net), and have very uniform connections between units. This is not typically true of living neural networks. Although "the brain is a collection of about 400 anatomically identifiable neural networks," the variation in wiring schemes (i.e. interconnections between neurons) is incredible, some being fantastically regular, others being completely random. In some sections there is relatively simple feedforward wiring, whereas other areas have absolutely incredible tangles of feedback schemes. Furthermore, the threshold functions of individual neurons depends on many different factors, including the type of neurotransmitter in the pre-synaptic neurons (i.e. how the different NT's from the different pre-synaptic neurons affect the post-synaptic neuron), the type of post-synaptic neuron (i.e. how the ion channels in the post-synaptic neuron interact, the different types of signal propagation along the dendrites, the different cell thresholds that cause the action potential, etc.), the manner in which excess neurotransmitter is broken down around in the synapse, how different hormones affect the neurons, etcetera ad nauseum! And that is just the beginning. As one might imagine, the differences are incredible. So I cannot see how one can derive that the brain is operating as a computer based on the insights that human brains are composed of living neural networks that resemble artificial neural nets (on the simplest level). In reality the two types of neural networks (artificial and living) are quite dissimilar, with the human brain having living networks that are orders of magnitude more complex than artificial ones. Please do not read into this that I think artificial neural nets are useless pieces of research, because I do think they have brought many promising results to AI, science, and humanity. I am still very interested in them. But after learning about real, living neural networks, I have discovered that we still have a long long way to go before artificial neural networks become as powerful (and complex) as human brains. And I do not think that the human brain is a computer, but I do think that machine intelligence is possible. >--Barry Kort .oO Chris Oo. -- Christopher Lishka (608)262-4485 Wisconsin State Laboratory of Hygiene lishka@uwslh.slh.wisc.edu Data Processing Department ...!uunet!uwvax!uwslh!lishka "... This week, Shane and Rebecca meet some squirrelly country boys. Steve and Kayla grow further apart, while Cal and Kim come closer to finding Arthur. Justin and Adrienne declare war." -- from TV Guide's "Soap Opera Guide"