X-Message-Number: 9306 Subject: misunderstood models Date: Tue, 17 Mar 1998 15:25:25 -0500 From: "Perry E. Metzger" <> > From: Thomas Donaldson <> > Date: Tue, 17 Mar 1998 00:17:42 -0800 (PST) > > I will add, though, that since real Turing machines are impossible (because > they have an infinite tape) it seems odd to argue that a machine able to > mark down real numbers (ie. infinite decimal numbers) or have components > a real number distance apart (such as pi or sqrt(2)) should be dismissed > merely because infinite decimal numbers are impossible. Turing machines don't actually possess infinite tapes. Permit me to explain what I mean by that. When we say a Turing machine has an infinite tape, we are really saying that it can use a tape of arbitrary length. A Turing Machine computation that is in any way bounded in time will of necessity use only a finite amount of tape -- the "Infinite" constraint is there to permit it to use a arbitrary amount of storage, as no program could actually access infinite storage in finite time. The human brain or your desktop computer do not actually possess infinite storage -- they are in fact a weaker construction than a Turing machine called a Finite State Automaton, because they are bounded. We us the "Turing Machine" as a model of computation because it is convenient and not because anyone actually would want to build one. It is a maximal model in the sense that it is believed that no real physical system could compute anything that a Turing Machine could not compute. Any real machine or construction is necessarily more constrained. Let us not forget the point of this discussion: we are debating a very narrow topic, which is "can a digital computer simulate a human mind". To argue against that successfully, Mr. Donaldson and Mr. Ettinger are going to have to show that the neural networks making up the brain are somehow capable of producing non-Turing computable results. It appears that real neurons operating in real brains do not use magical powers but in fact are based on electrochemical reactions, and are thus in principle straightforward to simulate. Given the practical engineering knowledge we have gained over the last 50 years of working with stored program digital computers, there does not appear to be evidence that, in the long run, computers will not possess adequite power to simulate even systems as complex as human brains. Perry Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=9306