[extropy-chat] A quick AGI question

Colin Geoffrey Hales c.hales at pgrad.unimelb.edu.au
Fri Nov 10 14:44:09 UTC 2006


> Hi Colin,
>
>   Thank you for responding. Even though I have only the most preliminary
> and limited understanding of AI, the information you have provided
certainly sounds good to me (FWIW) :-)  But, I think that I managed to
articulate my question in a sufficiently obscure way, that I think it's
intent may have been lost. I'll try to rearrange the question the best I
can, with the limited understanding that I have.
>
>   It seems to me that the human brain as a physical object, can and does
> produce intelligence and consciousness without what is commonly
> understood to be software. "Software" is just an invented human noun,
typically used to describe computer code that may be recorded in any
number of ways (eg. on a hard drive, within a program, or scribbled with
crayon on a piece of paper). IOW, the algorithms that produce human
intelligence seem to be supplied solely by the physical arrangement of the
hardware of the brain. So my question is: Is the premise behind AGI (or
really any software program, I suppose) that the active (running) software
*pre-specifies* the physical arrangement of the hardware (eg. by
specifying which transistors are active at what time) - and that this
newly pre-specified hardware arrangement is what then "produces" the mind,
moment by moment? IOW, isn't it ultimately the *hardware* that produces
the mind, even though it is the software which is dictating the
>  physical arrangement of the hardware, moment by moment?
>
>   Sorry if this question is still clear as mud, I'm having a hard time
of
> it.
>
>   Best Wishes,
>
>   Jeffrey Herrlich
>

OK. These sorts of questions are good. What they are not is quick! :-)

I need all the practice I can get at articulation of ideas. It's one thing
to understand, another to explain and yet another to 'build one'. The
computer metaphor for the brain is very pervasive and can be very useful
in one context and quite misleading in another.

I think I'm getting closer to your question, which relates to the
difference between abstracted computation and that 'computation' done by
brain material and what may be different between the two as chunks of the
universe doing something.

Firstly is a matter of "physical arrangement of hardware" as determined by
software. Peripheral devices are irrelevant in this. It is the central
processor which matters. The hardware is not changing physically in the
sense of matter creation/deletion or even rearrangement of atoms. All
these things are fixed in modern semiconductor technology. Sub-atomic
particles get maniulated (electrons/holes). Rather the hardware is
changing 'electrical state'. In a computer the specific electrical state
of the processor and the logical state of the program are not directly
related in any predefined systematic way. For example a subroutine may be
invoked many times, each in a different area of memory. One invocation may
be in static RAM, another in dynamic RAM. (There is the role of the
operating system inserted in all of this, too.) The electrical state of
the hardware in each case is completely different in physical location and
temporal behaviour (eg speed)....but it's all the same program. A
software-based (abstracted) AGI programmer would claim that the program is
determining the 'mind' thus created, not the hardware.

Secondly is the notion of a program in the case of brain material. Which
is where the computer metaphor does us a disservice. I now see that Robert
Bradbury has chimed in on this! I'll try not to double up.

The brain is a self constructed learning engine. Neurogenesis - the
initial creation of the brain (nervous system), its basic structure and
certain innate 'knowledge' (such as how to breath, for example) are the
result of 'programming' in the form of DNA and the interaction this has
throughout development through the role of expressed genes, intimately
co-ordinated
until...voila....we have something capable of learning how to learn
things. The situated brain is now stimulated with sensory feeds and the
stimulation
participates in the development of faculties needed to learn (learning how
to learn).

Having finished learning how to learn (development...around age 25 for
humans, when the great neural dieback has completed) what you end up with
is a machine that is innately programmed to learn. The capacity to learn
is hardwired by the structural/cellular/molecular biological reality of
the situation. "That which is learned" is represented in the
moment-to-moment reconfiguration of the existing structure of the brain in
the sense of connections, not cellular birth/death. A very complex
molecular dance indeed.

So there is no 'program' in a brain except it's innate learning capacity.
A perfectly healthy brain can, when exposed to the appropriate stimulus,
come to learn (believe) in diametrically opposed views, both views are
valid beliefs. The process of learning is the same. For example I could
believe in the great galactic pumpkin as the source of all natural laws.
You can believe the tooth fairy does that job. Both beliefs are rational
in the sense that they have been acquired by a healthy brain (healthy
learning biochemical algorithm).

So, whilst conceptually the computer metaphor is useful in some contexts,
computer and brain are actually very very different critters.

How are we doing thus far?

regards,

Colin





More information about the extropy-chat mailing list