belmonte@svax.cs.cornell.edu (Matthew Belmonte) (09/10/86)
In article <134@omepd>, max@omepd.UUCP writes: > >.. The difference -- or at least a possible difference > > -- is that digital circuits > >are designed to correct for small random fluctuations; the output of a TTL > >chip is virtually uncorrelated with the fifteenth significant digit of its > >input. I have seen no evidence as to whether this is true of human neurons, > >or whether they can tend to amplify these small fluctuations. > > The presence of theories accurately > modelling the behavior of neurons would seem to indicate that the large > scale behavior of neurons does NOT depend on QM events. > > So, if this 'information' is being used, it is being sent out > along the axon in equally small fluctuations. Postulating that > huge amounts of information are being transmitted all over the brain by > immeasurable fluctuations in neuron activity seems a little implausible, > and could be characterized as wild speculation. > I am certainly not well-versed in biology, but... It was my understanding that a neuron fires iff the voltage it feels from a neighbouring axon across a synapse attains a certain threshold level (assuming, of course, that it has had enough time since the last firing to recharge). so, neurons are basically digital devices (the biological equivalent of the electrical engineer's TTL buffer). It doesn't seem that small-scale variations in charge or electric field could cause changes in the behaviour of the brain. Certainly quantum effects would not induce any nondeterminism (unless there were an Improbability Drive -equipped ship passing through the universe at the time :-). Suppose for the moment that the above reasoning is faulty, and that random effects really do influence the behaviour of individual neurons. This does not imply that random events influence the behaviour of the brain as a whole. The brain must be redundant somehow; otherwise I'd be in danger of losing a facility or memory the next time I go drinking & cause the death of some critical neuron. -- Matthew Belmonte ARPA: <belmonte@rocky.cs.cornell.edu> <belmonte@svax.cs.cornell.edu> BITNET: <d25y@cornella> UUCP: ..!decvax!duke!duknbsr!mkb
friesen@psivax.UUCP (Stanley Friesen) (09/11/86)
In article <15600@ucbvax.BERKELEY.EDU> (David desJardins) writes: > *Everything* exhibits QM indeterminacy. For a neuron, I am sure that >if you measure the exact instant that it fires (say, as determined by the >point at which its output first exceeds a threshold value), you will find >that the fifteenth significant digit is as random as you could want. Quite probably, but I doubt this is of any biological significance(for more detail, see below). > The same is true of a TTL chip -- if you measure the fifteenth significant >digit of its output voltage, you will find it to be completely random. The >difference -- or at least a possible difference -- is that digital circuits >are designed to correct for small random fluctuations; the output of a TTL >chip is virtually uncorrelated with the fifteenth significant digit of its >input. I have seen no evidence as to whether this is true of human neurons, >or whether they can tend to amplify these small fluctuations. Just because the brain does not use exactly the same methods as electric circuits in smoothing out minor fluctuatiations does not mean that it doesn't do so. In fact it has several different mechanisms, operating at different levels, which tend to damp random fluctuation, except where it is useful. First is something very like what curcuits use, a threshhold system. Discharge of a neuron is triggered by an incoming signal which exceds a certain threshhold. In most cases this lower limit is considerably larger than the signal generated by a single input neuron. Second, the coding used by the nervous system is such that a single discharge has little or no meaning, it is the *rate* of discharge that is meaningful. Small scale variations in the timing of an individual spike simply have no effect on the averaged rate of discharge of a spike train. Third, the nervous system is highly parallel, with considerable overlap between adjacent data streams. Thus most significant results are based on a consensus of several semi-independent "computations", much like the use of triple redundency in "reliable" computer systems. So the brain in fact has considerable means of suppressing unwanted randomness. > And, in particular, since we don't know how the neurons are organized, >we don't know how subsequent neurons will be affected by random fluctuations >in their inputs. You would be suprised at how much we *do* know about how neurons are organized, at least in certain parts of the brain such as the Cerebellum, the Visual Cortex, and the Sensori-Motor Cortex. One of the main research problems now is the *intermediate* level of organisation, between the large scale organization of the neuro-anatomist and the small scale organization of the local processing complex. > As has been noted, carrying information is different from computational >power. A machine with a RNG is in certain ways more powerful than one >without. It is certainly a better poker player! Here I agree with you, the brain needs, and almost certainly has, a randomness generator, perhaps more than one. I rather doubt this takes the form of a random *number* generator though. Of course any randomness generator can be constructed from a RNG, so this is not a computationally significant point. --- Sarima (Stanley Friesen) UUCP: {ttidca|ihnp4|sdcrdcf|quad1|nrcvax|bellcore|logico}!psivax!friesen ARPA: ??
desj@brahms.BERKELEY.EDU (David desJardins) (09/14/86)
In article <1433@psivax.UUCP> friesen@psivax.UUCP (Stanley Friesen) writes: >> As has been noted, carrying information is different from computational >>power. A machine with a RNG is in certain ways more powerful than one >>without. It is certainly a better poker player! > > Here I agree with you, the brain needs, and almost certainly >has, a randomness generator, perhaps more than one. I rather doubt >this takes the form of a random *number* generator though. Of course >any randomness generator can be constructed from a RNG, so this is not >a computationally significant point. This is the only point I was trying to make, so I think we are in essential agreement. I certainly did not mean to claim that *every* part of the brain is highly nondeterministic (obviously the firing of the optic neurons is almost completely determined by the detection of photons by the retina), only that *some* parts of the brain can reason- ably be expected to behave randomly/nondeterministically/chaotically (chaos + small randomness = total randomness). -- David desJardins
jon@amc.UUCP (Jon Mandrell) (09/14/86)
-- Jon Mandrell (ihnp4!uw-beaver!tikal!amc!jon) Applied Microsystems Corp. "flames >& /dev/null" - me