machine brains

Malcolm McMahon malcolm at pigsty.demon.co.uk
Tue Mar 9 03:09:58 EST 1999


On Mon, 08 Mar 1999 14:08:01 -0500, Michael Edelman <mje at mich.com>
wrote:

>
>
>Malcolm McMahon wrote:
>
>> >> But it _is_ pretty clear what constitutes wiring and what constitute
>> >> transient electrical states in that wiring.
>> >
>> >But that still doesn't tell us what's program and what's data.
>>
>> Indeed, as a programmer I don't believe that there is any firm
>> distinction. Data is more volatile. That's about it.
>
>Unless you ROM the code, even that is up for debate.
>

Data too can be in ROM.

>> > Changes in wiring may be
>> >actual axonal growth, or it may be long-term potentiation of synapses. And axonal growth
>> >may be repair, not memory change.
>>
>> True, and dendritic growth is another aspect. In fact the latest news is
>> that new neurons actually grow in mature rats' brains contrary to
>> previous orthodoxy. But I don't have much doubt that a lot of these
>> changes are related to learning.
>
>But since "learning" can mean any change in the representational structure of the brain, that
>may not buy you anything.

Even if not all changes are construtive learning constructive learning
is undoubtably such changes.


> There's a lot of integration of memory that takes place during REM
>sleep,

Hang on, last I heard that's not been proved. There's an idea that
dreaming is a byproduct of a process of random stimulation that helps to
tidy up and increase the separation between neural nets but LTM storage
clearly happens all the time, at least whenever the hippocampus fires.

> and if that resulted in structural change would you call that learning?

Yes.

>
>> I suspect there's something ressembling an evolutionary process going on
>> in the CNS with neurons, dendrites, axions and synapses all forming more
>> of less at random all the time and then dying back if they fail to
>> justify their existance.
>
>That would be very energy inefficient. Neurons don't grow at random. There needs to be some sort
>of activity or chemical trigger.
>

It may be the only way to go. It is known that most young neurons which
are produced during the growth of an embryo self-destruct. There may be
no practical alternative to a kind of trial and error process. There are
clearly ways such a process could be improved beyond pure randomness,
for example new growth being encouraged by high levels of neural
activity. It needn't be that energy innefficient. The material from
cells that suicide is available for other cells.

>> >ECT typically affects LTM as well, and I don't think it's accurate to say it "completely
>> >scrambles the electrical state of the brain". It doesn't scramble it at all- rather, it
>> >synchronizes it, much as occurs in a seizure- which is where the idea for ECT came from.
>>
>> I think it's probably pretty much like when your PC locks up and you
>> switch it off for long enough for the capaciters to discharge, then
>> start it up again.
>
>Mm. Not quite.
>
>In a seizure, you get a situtaiton where a pattern begins in one area and radiates outward from
>that area, with the activation spreading outwards until a large portion of the neurons are
>firing in syncrhonization. This goes on until limited by depletion of neurotransmitters, lack of
>oxygen or some other limiting activity.
>

Yup, but the effect is the same in that either will disrupt any volatile
data in the system which will have to start with only non-volatile data.
I'd speculate that certain pathological mental conditions might be the
equivalent of a computer being stuck in a loop, or blocked by some kind
of mutual lockout.

>>
>> I certainly don't dismiss its existance, only that it can be observed
>> other than subjectively.
>
> So the quesiton is whether we can do science on conciousness. I think we can.

We may be able to look at the interface between consciousness and the
rest of the universe and learn a lot about how consciousness behaves.
This doesn't necessarilly tell us anything about what consciousness is,
though it might prove very usefull.

>> .
>> >Quarks.
>>
>> Well, no, not leptons but that's not important.
>>
>> > And quarks in turn may be made of simpler constructs.
>>
>> Again, in that case the simpler constructs are the elementary particles
>> and quarks are not.
>>
>> > All of the physical
>> >world may in turn be a complex wave superimposed on the background of space and
>> >time itself. The jury is still out on this one ;-)
>>
>> In which case the waves and space and time itself are the elementary
>> phenomina. It doesn't matter. Reductionism can only take you so far then
>> you have to accept you're reached the elementary.
>
>Yeah, but now we're arguing metaphors. Conciousness and mind are not simple constructs
>like leptons. You can't encapsulate everything about a particular conciousness with a
>couple of (relatively) simple equations.

Firstly as below I suspect consciousness needn't be complex (and I'm
certainly not claiming the mind is elementary in this sense).

But then who's to say all elementary phenonmina have to be simple anyway
(not that an electron's behaviour is all that simple).

>> That seems to me more like a leap of faith than an explanation.
>
>Rather an extension from observations of a great number of phenomena in the physical
>world. Taking the favorite examples of the ant hill or the bee hive or the slime
>mold,  we see how simple, predictable behaviors in the constituernt organisms react
>with each other and with the environment and produce what looks like intentional
>behavior. This seems liek a good model for the brain.
>

Yes, and I see the mind as a process occuring in the brain. I don't,
however, see consciousness as anything like mind.

>> Yes, I think the difference between reflex and thought is only one of
>> complexity.
>
>Except that thought is not just the sum total of conditioned responses of the cells.
>The critical aspect- and I'm basically arguing for the connectionist stance here- is
>the interaction. The number of possible states and possible arrangements gets really
>big really fast. The number of possible states of the brain is, well, a huge number.
>You couldn't enumerate them in the known lifetime of the universe.
>

Yes, it's a very wide order of difference of complexity, but it's still
a difference of degree.

>>
>> No, I don't think you see consciousnes in that. What you see is only
>> egos, which is to say collections of beliefs and habits.
>
>I would argue that you percieve an entity that appears to be not unlike yourself- and
>since you believe that you have conciousness, you assume that the entity at the other
>keyboard does as well, which is a reasonable assumption unless you want to argue for
>solipsism.
>

Exactly. You percieve the actions of a mind like yours and so you
_assume_ the presence of a consciousness like yours. You don't see
consciousness, or necessarilly any result of consciousness.

We look at our fellow humans (at least) and notice they ressemble us in
many ways. We therefore assume they also ressemble us in the matter of
being conscious.

>>
>> Rather, what it does is to deal with those that say that nothing
>> constructed of known particles can exist.
>
>Are you taking a sort of vitalist stance here- that conciousness is a non-physical
>substance? I'm not entirely clear as to your argument on this point.

What I'm saying is that something can be physical, and real without
being composed of material known to physics. This is, I guess, a
vitalist position in some ways (though there's no proof that all life is
conscious or that all consciousness exists in living things). 


>
>> > And anyways,
>> >conciousness is far too complex to compare it to an electron ;-)
>>
>> Now that's an interesting aspect. If you separate consciousness from
>> mind, and ego then why does it need to be complex?
>
>Re my earlier argument, which relies on a definition of complexity: How much
>information does it take to describe? The amount of information needed to describe a
>unique conciousness is far greater than that needed to describe a unique electron.
>

Not necessarilly because the detail you're talking about is the detail
fo the ego, not of the conscious self.





More information about the Bioforum mailing list