mangoe@umcp-cs.UUCP (Charley Wingate) (10/27/85)
There are a number of reasons why I doubt that the mind is in fact like a turing machine. As James Lewis pointed out, neurons are essentially analog devices; they typically respond to input *levels* rather than to descrete stimulae. If this holds up in general, it means that the state space of the mind is essentially continuous. THe brain also has a lot of analog inputs, posing similar problems with respect to the "tape". At the neural level, one typically sees random behavior. Increasing stimulae tend to increase or decrease the rate or probability of firing, but the action is distinctly unlike the firing structure of a computer. State transitions would seem therefore to be probabilistic. Lastly, it's certainly clear that we cannot now model even moderately small portions of the mind through computers. I think it is reasonable to ask those who wish to assert the turing machine-ness of the mind need to show some method by which the mind can be translated into an equivalent turing machine, even if this translation is computationally infeasible (which is indeed likely). Without such an algorithm, I think there is reasonable cause not to accept the hypothesis. Charley Wingate
greg@hwcs.UUCP (Greg Michaelson) (10/29/85)
> There are a number of reasons why I doubt that the mind is in fact like a > turing machine. > > Lastly, it's certainly clear that we cannot now model even moderately small > portions of the mind through computers. I think it is reasonable to ask > those who wish to assert the turing machine-ness of the mind need to show > some method by which the mind can be translated into an equivalent turing > machine, even if this translation is computationally infeasible (which is > indeed likely). Without such an algorithm, I think there is reasonable > cause not to accept the hypothesis. > > Charley Wingate We cannot now do X THEREFORE we cannot ever do X where X = build a heavier than air flying machine = transmute one substance into another = model brain behaviour with a computer etc etc etc
mangoe@umcp-cs.UUCP (Charley Wingate) (10/31/85)
In article <667@hwcs.UUCP> greg@hwcs.UUCP (Greg Michaelson) writes: >> Lastly, it's certainly clear that we cannot now model even moderately small >> portions of the mind through computers. I think it is reasonable to ask >> those who wish to assert the turing machine-ness of the mind need to show >> some method by which the mind can be translated into an equivalent turing >> machine, even if this translation is computationally infeasible (which is >> indeed likely). Without such an algorithm, I think there is reasonable >> cause not to accept the hypothesis. >We cannot now do X THEREFORE we cannot ever do X > where X = build a heavier than air flying machine > = transmute one substance into another > = model brain behaviour with a computer etc etc etc Well, the correct analogy in the first case is X = Build a Flying machine with flapping wings and in the second case X = Transmute a substance using alchemy which fit well with the third X = Model the brain with a VonNeuman machine I'm not arguing that we can't model the brain with a computer. I'm just saying that the efforts of AI researchers tend to indicate that such computers aren't likely to be like today's machines. In principle, for instance, we could build something which had lots of little chips, one for each neuron. It's also important to note that this is not merely a technological question: it is also a statement about the nature of existing natural technology. One ordinarily expects such hypotheses to make experimental predictions which are then put to the test. Instead, it is believed in with a kind of religious fervor. It's perfectly find to persue this hypothesis further. But there's no reason for anyone to believe to be a truth. Charley Wingate
root@bu-cs.UUCP (Barry Shein) (11/01/85)
There are only three arguments that can feasibly disprove that the mind cannot be simulated by a computer: A. Semantic: If there were a computer which simulated the mind, we would not call it a computer (quite literally this is possible, consider for example that we needed to add biological components to accomplish the job, such as enzyme switches, this would no doubt cause semantic controversy.) B. Religious: This rests on attributing capabilities of the mind to non-physical components of a magical nature that, by their very definition, can not be simulated by physical systems. C. Existential: We could not prove that any given computer is simulating the mind, although a few clever arguments have been put forth for tests (Turing: if an observer cannot distinguish the machine from the human then it must be equivalent) these have generally been shown to be unsatisfactory primarily due to their lack of rigor (eg. who is to be chosen as the observer?) The problem of proving that a computer is simulating the mind I will claim is likely equivalent to simulating a mind. If we cannot prove something exists, then it does not exist. The 'proof' of this claim by me is that the mind is a finite object contained within the head of a human. It is a physical object composed of physical materials. If something is physical and finite then it should be possible to simulate its behavior -at least in theory- (eg. the entire universe may well be physical and finite, but is not very simulatable in practice.) One suspects that any counter-argument resting solely on the magnitude of the problem is only defining the problem, not refuting it's solvability. I claim the only counter-arguments to this fall into one of the above three categories and that none of these categories is a satisfactory proof of anything, just a rhetorical subterfuge. -Barry Shein, Boston University
cccjohn@ucdavis.UUCP (John Carlson) (11/01/85)
*** LINES FOR SALE: 50 CENTS WORTH FEEDS FAMILY OF FOUR *** In article <1996@umcp-cs.UUCP> Charley Wingate writes: > Lastly, it's certainly clear that we cannot now model even moderately small > portions of the mind through computers. I think it is reasonable to ask > those who wish to assert the turing machine-ness of the mind need to show > some method by which the mind can be translated into an equivalent turing > machine, even if this translation is computationally infeasible (which is > indeed likely). Without such an algorithm, I think there is reasonable > cause not to accept the hypothesis. Later, in article <2031@umcp-cs.UUCP> he adds: > > I'm not arguing that we can't model the brain with a computer. I'm just > saying that the efforts of AI researchers tend to indicate that such > computers aren't likely to be like today's machines. In principle, for > instance, we could build something which had lots of little chips, one for > each neuron. It's also important to note that this is not merely a > technological question: it is also a statement about the nature of existing > natural technology. One ordinarily expects such hypotheses to make > experimental predictions which are then put to the test. Instead, it is > believed in with a kind of religious fervor. > > It's perfectly find to persue this hypothesis further. But there's no > reason for anyone to believe to be a truth. 1) Assume you could design a Turing-like machine equivalent yourself. 2) Then you could comprehend all of this machine's actions, because you would know all of it's inputs and outputs. 3) Then you could comprehend all of your actions. 4) Knowing all of the machine's inputs and outputs is equivalent knowing the universe. 5) Then you could comprehend the whole universe. 6) I conclude that it would be easiest to model a human being with an analog machine. 7) Try replacing "you" with "we", "yourself" with "to a human", and "your" with "a human's". Here's some questions: 1) Are all inputs and outputs equivalent to the universe? 2) Can we make something we will never comprehend, that is, a higher intelligence? John Carlson
lambert@boring.UUCP (11/02/85)
In article <212@ucdavis.UUCP> cccjohn@ucdavis.UUCP (John Carlson) asks: > 2) Can we make something we will never comprehend, that is, > a higher intelligence? Parents can make something they will never comprehend, namely their children. Now I understand why human intelligence is ever-increasing. (Sorry, couldn't resist that.) Seriously, the incomprehensibility of a human-made formal system does not imply it is "higher". -- Lambert Meertens ...!{seismo,okstate,garfield,decvax,philabs}!lambert@mcvax.UUCP CWI (Centre for Mathematics and Computer Science), Amsterdam
rlr@pyuxd.UUCP (Rich Rosen) (11/02/85)
>>We cannot now do X THEREFORE we cannot ever do X >> where X = build a heavier than air flying machine >> = transmute one substance into another >> = model brain behaviour with a computer etc etc etc > Well, the correct analogy in the first case is > X = Build a Flying machine with flapping wings > and in the second case > X = Transmute a substance using alchemy > which fit well with the third > X = Model the brain with a VonNeuman machine > > I'm not arguing that we can't model the brain with a computer. [WINGATE] The argument is with those who insist that we cannot model the brain with a machine at all. -- "Mrs. Peel, we're needed..." Rich Rosen ihnp4!pyuxd!rlr
jbuck@epicen.UUCP (Joe Buck) (11/03/85)
> From: cccjohn@ucdavis.UUCP (John Carlson) > 1) Assume you could design a Turing-like machine equivalent to > yourself. > 2) Then you could comprehend all of this machine's actions, > because you would know all of it's inputs and outputs. > 3) Then you could comprehend all of your actions. Statement 2), even when applied to a much simpler system that a person (such as a theory of the natural numbers 0, 1, 2, ...), is what Godel disproved. That is, even though we write down exactly what the rules of arithmetic are, there are an infinity of statements that we can't determine the truth of. This is a common fallacy made by people who argue against machine intelligence: that knowing the inputs and the rules of a machine, you understand it completely and it can never surprise you. This just isn't so. If you personally don't suffer from this limitation Godel discovered (as some people like to argue) I have a few computer programs I'd like to have you debug. :-) -- Joe Buck | Entropic Processing, Inc. UUCP: {ucbvax,ihnp4}!dual!epicen!jbuck | 10011 N. Foothill Blvd. ARPA: dual!epicen!jbuck@BERKELEY.ARPA | Cupertino, CA 95014
greg@hwcs.UUCP (Greg Michaelson) (11/04/85)
> In article <667@hwcs.UUCP> greg@hwcs.UUCP (Greg Michaelson) writes: > > >> Lastly, it's certainly clear that we cannot now model even moderately small > >> portions of the mind through computers. I think it is reasonable to ask > >> those who wish to assert the turing machine-ness of the mind need to show > >> some method by which the mind can be translated into an equivalent turing > >> machine, even if this translation is computationally infeasible (which is > >> indeed likely). Without such an algorithm, I think there is reasonable > >> cause not to accept the hypothesis. > > >We cannot now do X THEREFORE we cannot ever do X > > where X = build a heavier than air flying machine > > = transmute one substance into another > > = model brain behaviour with a computer etc etc etc > > Well, the correct analogy in the first case is > X = Build a Flying machine with flapping wings Have you not seen the flying elastic powered plastic pigeons with flapping wings? > and in the second case > X = Transmute a substance using alchemy > which fit well with the third > X = Model the brain with a VonNeuman machine So VonNeuman technology = alchemy? Using current chemical/physical theory it can be proved that alchemical techniques cannot transmute substances. Can you provide an equivalent proof that VonN machines cannot be used to model the (admittedly vast) finite state machine inside human skulls? The form of argument is fallacious. I put in schematic form to try and this apparent. > I'm not arguing that we can't model the brain with a computer. I'm just > saying that the efforts of AI researchers tend to indicate that such > computers aren't likely to be like today's machines. In principle, for > instance, we could build something which had lots of little chips, one for > each neuron. It's also important to note that this is not merely a > technological question: it is also a statement about the nature of existing > natural technology. Why should 'natural technology' have any relevance for technology in general? > One ordinarily expects such hypotheses to make > experimental predictions which are then put to the test. Instead, it is > believed in with a kind of religious fervor. It is actually religion which is affronted by the suggestion that humans are no more than than protoplasmic automata. Just because people in AI make loony claims does not mean that the computer simulation of human behaviour is impossible. > It's perfectly find to persue this hypothesis further. But there's no > reason for anyone to believe to be a truth. How about some hypothesis to show that it can't be done? Men on the moon? Nonsense! Travel underwater? Balderdash! Destroy a city at the push of a button? Gad sir!
mangoe@umcp-cs.UUCP (Charley Wingate) (11/06/85)
In article <2012@pyuxd.UUCP> rlr@pyuxd.UUCP (Rich Rosen) writes: >> I'm not arguing that we can't model the brain with a computer. [WINGATE] >The argument is with those who insist that we cannot model the brain >with a machine at all. Is it? Or isn't it really an argument over the kinds of machines which offer the hope of success? Charley Wingate umcp-cs!mangoe
mangoe@umcp-cs.UUCP (Charley Wingate) (11/07/85)
In article <677@hwcs.UUCP> greg@hwcs.UUCP (Greg Michaelson) writes: >> Well, the correct analogy in the first case is >> X = Build a Flying machine with flapping wings >Have you not seen the flying elastic powered plastic pigeons with flapping >wings? Certainly, and those existed back before the Wright Bros. did their thing. No sign of man-sized versions, though. Anyway... >> and in the second case >> X = Transmute a substance using alchemy >> which fit well with the third >> X = Model the brain with a VonNeuman machine >So VonNeuman technology = alchemy? Using current chemical/physical theory >it can be proved that alchemical techniques cannot transmute substances. Can >you provide an equivalent proof that VonN machines cannot be used to model >the (admittedly vast) finite state machine inside human skulls? My point here was not VonNeuman machines CAN'T do it-- it's that there's a strong possibility that the V.N. archetecture is simply the wrong mindset from which to approach the problem, much as flapping wings and alchemy were to their problems. Too often the voice I hear from the AI-ists is "V.N. (or Parallel, or whatever-your-favorite-variation) is the only way we know to attack the problem, so we will assume that it is the correct way." The notion that the mind is a great state machine is, I would contend, dangerously close to that sort of thinking. It's conveniently unfalsifiable, it's patently unmodelable as it stands (2**(10**10) states!?!), and thus allows you to work indefinitely on the problem without the inconvenience of being put to the test. What I don't hear these people saying, though, is "What are we going to do if it turns out NOT to be like a giant state machine?" One of my professors the other night made the claim that everyone should be a programmer, because that's the only way they are going to get what they want done on a computer. He persisted in an analogy between computer programming and writing. My personal opinion is that this is going to acheive the same results as we commonly see with programmers writing manuals; they supposedly know how to write, but they aren't really competent to write effectively on any large scale. But this is a side issue. My sociological comment on this is that it illustrates the sort of messianic light which one commonly sees in the eyes of computer scientists these days. Programming will change everyone's way of life. AI will give us new electronic brains. It's in some respects similar to the situation at the beginning of serious investigation into HTA manned flight; plenty of people thought it was possible, but almost without exception they were wrong about how it would be brought to pass. Charley Wingate
stark@sbcs.UUCP (Eugene Stark) (11/08/85)
> > Well, the correct analogy in the first case is > > X = Build a Flying machine with flapping wings > > Have you not seen the flying elastic powered plastic pigeons with flapping > wings? > I recall recently seeing an article concerning a life-size (I can't remember exactly, but I am reasonably certain >10ft wingspan) mechanical Pterodactyl model, already demonstrated or to be demonstrated soon. The mechanical Pterodactyl propels itself in flight by *flapping its wings*. If I remember correctly, the model was developed by Paul McCready & company (of Gossamer {Condor, Albatross, ...} fame). Gene Stark
breuel@h-sc1.UUCP (thomas breuel) (11/09/85)
Below are the answers to two questions that 'biep@klipper.UUCP' asked. Altogether, you can make many comparisons between the brain and Turing machines, but such comparisons will not tell you much about either theoretical or practical limitations of the human brain. Thomas. ---------- ** Why is time complexity not a useful measure for comparing a Turing machine with a real life architecture? Turing machines are very nice devices for theoretical considerations. In a sense, they give the most believable and strict measure of computational complexity. For real life architectures, the theoretical benefits of a Turing machine are unimportant. You can't, for example do accesses to data on a Turing machine in less than O(n), whereas in real life, even on a serial architecture, you can do them in essentially constant time. ** Can you define 'Turing equivalent'? Something is Turing equivalent if it can simulate a Turing machine. Since a Turing machine does not have a limit on the amount of information that it can store, anything that can simulate a Turing machine can also not have a limit on the amount of information that it can store. The human mind probably has such a limit (judging from its architecture). Therefore, the human mind is probably not Turing equivalent.