Bill

The 33 MHz PC in my collection dates from roughly 1989. The 1986 IBM-type PC would be the 16 MHz 286. This had the computational power of roughly an IBM 360 Model 40, a huge mainframe of roughly 1968 vintage. We have yet to develop a good approach to 3-dimensional electronic circuits. We've done very well at shrinking them along two dimensions, but the third dimension seems to be mostly taken up with a heat sink, required to dissipate the enormous amounts of heat energy produced by the very inefficiencies of electronic switching circuits.

On the positive side, however, the number of transistors per square mm continues to increase (in accord with Moore's Law). The top-of-the-line Intel Itanium processor has 25 million transistors in the CPU and 300 million within its on-board cache memories. However, these numbers are difficult to convert into neurons because there isn't an adequate model of just how many transistors it takes to effectively model a neuron. Of course, electronic computers have a cycle time measured in GigaHertz, so you can make up for the lack of quantity through some sort of time-multiplexing. As you say: Lets suppose, for the sake of argument, that I can design a circuit that models the function of a human neuron in a space consisting of about 300,000 transistors. On a single chip containing 300 million transistors, I could then have 1,000 neurons. If I could model the neuron in 30,000 transistors, that number would (of course) increase to 10,000 neurons per chip. You would thus need an array of from 100,000 to 1,000,000 of those chips to get to the one billion neuron level. Using fairly standard packaging, I could probably mount 100 of those processor chips into a standard relay rack. So, I would need from 1,000 to 10,000 relay racks of neurons, using some sort of high-speed architecture to communicate (parallel gigabit ethernet channels is fairly cheap). Even the high end of this number (100 rows of racks, each with 100 racks per row) is very achievable with current technology. The typical large telecommunications switching site has more equipment than that would represent. And this is today.

Use Moore's Law for a decade and you get that down to just over three rows at 100 racks per row. Another decade reduces it to a total of about ten racks of equipment. Before the end of the third decade, you are at a single rack and shrinking further.

The above analysis presumes that a fairly large number of dedicated transistors (300,000) are necessary to model any given neuron. If it is one tenth of that amount (30,000 transistors), then we get to the single rack at the end of two decades of Moore's Law.

And the brain cycle rate of 40 Hz, versus the electronic processor cycle time (currently) of over 1,000,000,000 Hz, certainly demonstrates that there is a great deal of opportunity for time sharing the electronic circuits. Even assuming that inefficiency in making transitions requires 1,000 electronic cycles to emulate 40 brain cycles, you could still to a million-to-one reduction in hardware through time sharing. That would imply that the original 3,000 neuron chip, reduced to 1,000 neurons and set up for time sharing, would do the job of a whole billion human neurons.

=====

In other words, I don't see a technological barrier to be overcome here. What we have is a knowledge gap. We really don't know how human neurons actually work, so we are incapable of modeling them with any precision. If we knew how to model them, we could certainly construct the necessary hardware today, not five decades into our future!

== Bill

Bill

I think that subsequent discussion has effectively rebutted this misconception (being based, as it is, on a very-obsolete understanding of what computer technology is capable of). That is only true if you only ask those philosophers who believe that "mind" is a "substance" that is separate from "matter" (or something along the lines of mind not being "reducible" to matter). There are, of course, many philosophers who take the opposing point of view (the "materialists"). You might try perusing the <a href="http://www.artsci.wustl.edu/~philos/MindDict/index.html" target="_blank">Dictionary of Philosophy of Mind</a> for a plethora of references to both camps. Frankly, it is my understanding that the "materialists" had something of the upper hand in Philosophy of Mind circles these days. For example, on <a href="http://artsci.wustl.edu/~philos/MindDict/dualism.html" target="_blank">THIS PAGE</a> it says: "Arguments against dualism have been provided on the basis of both empirical evidence and on philosophical grounds, and clearly express the predominant view (Dennett, Damasio, Churchland). However, a number of modern philosophers of mind, though in the minority, have come to the defense of dualism (e.g. Hart)." This isn't my field, Ierrellus, but in general, the better the model, the better the results. It is my general understanding that current AI neural models are extremely crude, and thus they could benefit from expansion in any number of directions.

== Bill

John Page

Used to be that biologists thought the grey stuff in the head was useless junk! Perhaps 97% of the Columbia Encyclopedia is junk? It's figuring out which 97% is the problem.

Ierrellus

Greetings All!

excreationist, please forgive me for addressing you as Walrus. I had not yet consumed my second cup of coffee. And thank you for the reference to Piaget. It is most useful.

John, "fartility"! LOL, I deserved that. But you must admit I wheedled out of the non-sequitur not too shabbily. I will try a better response after trecking to the local library.

Bill, amazing material presented in your post. Thanks for reference to The Dictionary of Philosophy of the Mind. Will comment later.

Ierrellus

Ierrellus

Our library here is deficient of books recommended here in these posts. I did locate THE UNDISCOVERED MIND by John Horgan. The author claims to explore the limits of neuroscience.

IMO on criticisms raised in our discussions: The cell may have functions not directly related to genetic input; however, the cell is a genetic construction.

In the example of identical twins one mother and one father must provide chromosomes for two offspring. It is only reasonable that the two offspring will be similar or identical in their dispositions regardless of disperate environments. Their adaptations to each environment will still be similar.


In my studies I have found examples of adults with memories of the foetal condition. Most of these examples are anecdotal. Bill, does your wife or you know of hard evidence of foetal thought?

Restatement of theory in light of good responses--Genetic activity emerges as tools for reason in mental content. For example, geometry was experential before Euclid.


Ierrellus

John Page

Ierrellus:

Have you looked at ancestral memory at all? I remember reading a book by an American dentists who hypnotized people (apparently is was part of a dentist's training way back) who sometimes came through with previous life stuff.

I tried a search for the book on the 'net but couldn't find it. If there are ancestral experiences available to the mind, I suppose the mechanism would have to be genetic or fraud.

Cheers!

Imhotehp

Hi bill.

this is sorta off topic and everything, but could you describe what exactly E=mc^2 is to me? Right now I think it is supposed to describe how much energy you get when you convert matter into energy.

Thank you in advance.

Ierrellus

John:

Yes, I've checked out ancestral memory in Jung, et al. In my youth the popular book on reincarnation was The Search For Bridey Murphey. Reincarnation is not only a tenet of the oldest religions, it is still a popular belief among most people of the world. What stops me from any public assumptions in that area is the scoff and scorn many philosophic minds place on such phenomena as reincarnation, NDEs, OBEs and even
Jung's theory of archetypes.

Since it appears to me that Piaget worked on ideas of post-natal evolution of the brain, my best bet would be to consider all evidence of prenatal "mental" functions. So the first step is to explore foetal "mind".

Ierrellus

Ierrellus

Bill:

On neurons--all seem to fit categories of signal, transmittal, and response. Those that promote motor activity are of the simplest level of this process. Those that promote higher cognitive functions apparently employ quantitive and qualitative interpretations of both signal and response. These adjustments are performed not only by neurotransmitters in the neural net, but by which areas of the brain the signal transverses. The confusion in tracking signal/response comes from multiple functions of brain areas. The "back-up" systems for intensity of signal,inhibition or modification are in several brain locations.


Ierrellus

[ April 02, 2002: Message edited by: Ierrellus ]</p>

snatchbalance

<a href="http://www.hosppract.com/issues/1999/07/gershon.htm" target="_blank">enteric nervous system</a>

I thought this was interesting, and maybe even pertinent. It seems there are more brains than the one in your head.

I know that my "gut" feelings often inform me better than my ideas.

Snatchbalance