Principles of neural sciences

[Hoda]

By kandel and Shwartz.. the best book about neurosciences i've read so far.

[austintorn@aol.com]

Hi Hoda,

Neuroscience has been telling us a lot about ourselves lately. What are some of the good parts of the book?

[Summaria]

Any biologists here? Cognitives?

[SteveA]

I've got some familiarity with artificial neural networks, artificial intelligence, genetic algorithms and machine learning ... not much with the biological arena, though there are ways in which most any area of computation could be generalized as something capable of "Universal Computation". The main differences then between specific forms of computation are in terms of efficiency and qualitative differences in input/output representations.

A simple representation of computation is just:

input->transformation->output

But in terms of a TOE there's a lot more to consider as to how a process with such apparent simplicity could exist and whether or not there would even be a way to witness this directly in a fundamental form.

If we assume experiences and knowledge etc. are the result of some "output" of such a process, then that implies the input and transformation would not be 'visible' to such knowledge and experiences.

If we're to remove unknowns, then it would appear the brain, as some physically witnessable and comprehensible object couldn't be considered the fundamental source of such cognitive abilities and that the mind would exist at least partly independent of a brain.

Another consideration would be that for continuity across time to exist, there should be feedback in this and elements of the "output" would also be equivalent to inputs and in that case, it could be rewritten simply as:

input(s)->cognitive/perceptual memory

With the input as entropy/energy being accumulated in a (perceptual) memory.

Form/logic/space/dimensions could appear as computational components of something that is fundamentally a record keeper ... a good question in that case would be over what specific properties the "input(s)" has/have.

[Summaria]

Subconscious not visible to thought.

What unknowns to remove?

[SteveA]

Subconscious not visible to thought.

What unknowns to remove?

Well there's one unknown that appears to be redundant. You can compress all unknowns into a single unknown present at every moment in time.

I don't think there would be an ability to fundamentally distinguish how many unknowns influence things.

For example, if we say:

x=y

but neither x nor y is known, then they're redundant. Either x or y describes the same thing ... too many unknowns and not enough constraints.

Similiarly this tells us little:

x+y=5

We should treat (x+y) as a single quantity in that case as specifically 5, unless there's some other manner to determine a relationship between x and y.

So there should be some fundamental unit of time that contains only a single unknown at every moment.

[Summaria]

Planck? But you may have lost me.

[SteveA]

Actually that's a great way to look at it. Yes, what are the properties of a Planck unit?

My comments are pretty abstract, but it seems like if we're talking about fundamentals then we're bumping up against absolutes that can't be conveyed easily except for indirectly. You have to recognize them for yourself and with your comment it appears you already understand that.

Another way of looking at the cognitive picture is that if we're using such cognition to understand things but assume that some subconscious process is also involved beyond recognition, then we're creating a picture in which some unknown/unobservable is being "prefiltered" by some similarly unknown/observable subconscious process (at least if we assume our conversation here is not subconscious ), and this could be a redundant view - if we didn't know what the input was in the first place, then transforming it into some other unknown format doesn't "gain anything "tangible" ... on the other hand it does gives us a potential target to explore - for example, we could attempt to consciously understand what subconscious processes do, but recognize that any of that that comes into conscious recognition won't truly be something subconscious.

Basically, I'm trying to just point out that it appears the simplest form in which an unknown can exist is just as a singular influence and as another mathematical analogy, if we have some unknown function, f(), and we have an unknown variable x, then if we say:

y=f(x)

Where y is some present experience arising from the function, representative of subconscious influences and x, the present "objective" state of things, but if we can precisely define f() or x, then we might as well remove the x altogether and simply say:

y=f()

and contain the uncertainty of x within f(), or even simpler (why have two arbitrary ways of writing the same function? ... that's just another irrelevant unknown)

f()

So that's basically what I'm trying to point out in terms of cognition and one of the main reasons why I don't believe the brain can explain things completely is because of space. The body itself would appear to have no manner to perceive space outside of itself. There should be no way to see space between photoreceptors of the "real" eyes unless we're working with measurable phase shifts in space/time, but once again space and time aren't directly physically measurable either. We can measure between endpoints of a rule but not arbitrary endpoints in space and similarly there's no way to point to a span of time except in terms of intangible memory and perception.

I'm not trying to say these things aren't real aspects of experience, but that they can't be controllably conveyed physically. They can be shared, even physically remotely, but it doesn't appear conventional physical sciences defining a brain are going to be able to work in that arena, or at least not in a manner that appears provably true.

On the other hand, I hate to say things are impossible, but in terms of conventional forms of logic and science that I'm familiar with, it seems impossible to do.

[SteveA]

My comments there aren't particularly helpful in bringing out detail we can work with, but instead were mostly targeted at trying to show what appear to be realistic boundaries that would constrain what possible forms of cognition we can refer to with much of any precision or work with logically etc.

If we added back in an unknown, it should ideally be a singular unknown influence and I think time is a good candidate, in which case we could say that a presently perceived state is:

f(t)

Though recognize that the function itself could potentially be something from an unlimited number of possible functions. There are specific characteristics though that would appear to be required to persist from moment to moment though, and so it should be at least figuratively continuous in some respect and not be a "random" mapping.

Notice also that if time is unbounded, then there are similarly an infinite number of possible values for t, and if the "perceptual space" that we could distinguish between was limited to some finite number of possibilities, then potentially an infinite number of identical moments could exist alongside each other at the same "presently perceived" moment.

We don't need to have both f() and t be of potentially infinite values and there's even a problem with trying to use two separate and independent infinite quantities, so one should be related to and a function of the other. In this case, it would be good to decide which is the larger/driving quantity - units of "Planck volume" spacetime or the number of distinct possible states of experience that could be distinguished between?

It would appear that in order to arrange those correctly, some assumptions as to what perspective of the universe we're using would have to be made - are we seeing an "objective/external" version of spacetime or a "first person" perspective? Realistically, I think we could only expect to be able to work with the first person perspective (what other perspective should truly be available to anyone?) and in that case, it could be that we'd have to place the possibilities for f() as the larger and driving factor for t, which seems quite an unconventional view of time and hard to "wrap my head" around that one, yet time is an interesting subject and may not act in the manner it's often treated to be. It's the most commonly occurring noun in English and is used in a ton of diverse contexts.

Anyway, these seem to be some fundamental issues or even paradoxes (for example could we truly address some fundamental form of Planck time or is that an impossible, or at least unreliable abstraction?) involved in trying to work with cognition.

[Summaria]

We don't so much have to strain to see far out, but that it comes to us.