X-Message-Number: 14953 Date: Sat, 18 Nov 2000 21:18:50 -0500 From: Thomas Donaldson <> Subject: more on machines and brains Hi everybody! I note the reply of Joseph Kehoe to my question about whether brains can really be assumed to be computers. My problem is not so much with what he said as his justification for it: several issues need to be remembered. 1. The changing connections of our neurons are not arbitrary. (Remember that each neuron is a small computer). The same may be said of the use of any newly created neurons. Both result from the work a brain is doing AT THAT TIME. This raises some problems which bear on any proof that the machine counts as a Turing machine. The problems come from creation of NEW tapes as distinct from using farther and farther away parts of a single tape. That the machine is working in real time adds an important feature to this: if the space on the single tape is far enough away, it cannot reach it in time to do the needed calculations soon enough. Creation of new tapes does not have this problem. 2. The possibility of imitating an entire brain with a single processor still raises problems. A variety of our neurons actually respond to sensory input; many more are processing sensory input. If we had only one processor (with lots more circuitry for memory etc) the possibility of missing sensory input still remains; the same problem happens with PROCESSING sensory input. We can't just look the other way while we're doing something else. As I said originally, the fact that our brains work in real time is not just a minor issue, but important for understanding them. I would like Joseph Kehoe (or anyone else) to give a much moredetailed answer, quoting results from books or papers, on this question. I am willing to be quite patient while waiting for that answer. Over and above any answers to me, I understand that some computer scientists have imagined machines with special features (say the ability to compute with real numbers) which makes them not Turing at all. Though our brains aren't primarily designed by evolution for computing with numbers, such possibilities need considering. In terms of actual results in the literature, support for brains as Turing machines requires not just that a machine with many "neurons" can be reduced to a machine with one, but that machines with varying numbers CREATED IN RESPONSE TO A NEED FOR A COMPUTATION can be reduced to just one. The simple fact that brains are also real-time machines also must be dealt with: at some point a large enough number of neurons will create an object which cannot be imitated by any single processor IN PRINCIPLE, for the simple reason that no single processor could do the required work fast enough. The fact that we're processing our sensory input makes a difference here. AS I've said before, but wish to emphasize here, I am not arguing that a machine imitating human beings enough to be conscious etc just like us is impossible. I am raising the question of whether or not such a machine is a TURING MACHINE. Best wishes and long long life to all, Thomas Donaldson Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=14953