[extropy-chat] A quick AGI question

[Colin Geoffrey Hales]

    <3011.139.168.42.79.1163171298.squirrel at webmail.student.unimelb.edu.au>
          <20061110174758.GX6974 at leitl.org>
In-Reply-To: <20061110174758.GX6974 at leitl.org>

> On Sat, Nov 11, 2006 at 02:08:18AM +1100, Colin Geoffrey Hales wrote:
>> A quick post-script....
>> 1) The alteration of brain 'hardware' in the form of new
>> neurons/astrocytes does occur, but more in the role of repair than in
learning. It is only relatively recently that the details of
>
> Shape changes and their property changes (minute/hour/day) determine the
short-term (ms/s) processes. In practice there is no sharp
> distinction data/program/hardware nor clean layer separation
> as in human-designed systems.
>
>> neural/astrocyte cell genesis/apoptosis have emerged. The picture is
not
>> complete.
>
> The picture is is very fuzzy, and damn complicated.
>
>> 2) The electrical 'state' of the brain is a very complex 3D electrical
field pattern. It literally is the "mind" in the sense that the
>
> Spatiotemporal patterns of electric gradients are only a facet of the
entire process. There are chemical gradients, machinery modulation, gene
pattern activity changes, and similiar.
>
>> manipulation of it is what the brain is doing. This is the main
computer/brain distinction. In the computer the electrical field
pattern
>> is irrelevant.
>
> "The" computer doesn't exist. FPGA defines gate connectivity
> by SRAM cell state, and spintronics can define a logic element
> by a spin configuration. People will use whatever works best.
>
Hi!

>From Jeff's perspective... note the great variation in the range of views
on the role of hardware in conmputation and then in intelligence and
cognition... yours and Ben Zaiboc display one view of the relationship
between the 'computer' and 'computation'....I differ greatly...

....which is I would claim that an fpGA based AGI (I originally thought it
ould do!), for example, would have no 'mind' in the sense that we have.
Nor would it have intellect like ours. Nevertheless it could be a useful
device applied to something domain/habitat specific. The empirical
prediction would be fragility/maladaptation outside that domain/habitat. I
am quite close to mathematical proof of this (which is a kind of oxymoron,
but there you go) and my PhD project is computationally exploring the
mechanism details. In the end, special hardware is needed... there's a
huge background (3 years of analysis) to this... too much for here, now.

All in good time.

Meanwhile, another view of computation for Jeff:

Try this: pick up a nearby <object>. Hold it in your hand, look at it and
say "the universe has computed the <object>". That is the notion of
computation important to me. In humans, the universe has actually computed
a learning (= computing) machine called the brain and has made full use of
the innate circumstances of situatedness within the
universe-as-computation, not any abstraction constructed by the brain thus
reified in hardware or software........

So have we broken your brain yet, Jeff?

:-) It's so nice to be back in a place where people get this stuff. So
much of academia is lost in dogma and paradigmatic industrialised
mega-scale science...I'm finding it refreshing here....I've been in the
midst of wet neuroscience for several years... amazing. email forums full
of dogma, frightened to explore beyond the financially/professionally
threatening boundaries of the received view...Computation is the _last_
thing the bulk of mainstream neuroscientists think about. Other than
figuring out statistical significances of their data(is my N big
enough?)...These are the poor devils intent on figuring out the details of
molecular-pathway X, developmental issue Y details, pathology W
origins/progression mechanisms, knockout mouse Z gene expression etc....
important, support it, gotta be done...but sooooooo boring! And of course
they think I am from the planet zork, speaking in tongues. I seem to have
digressed somewhat. Ooops.

cheers,
Colin Hales