Newsgroups: comp.ai
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!nntp.sei.cmu.edu!news.psc.edu!hudson.lm.com!news.math.psu.edu!news.cac.psu.edu!newsserver.jvnc.net!newsserver2.jvnc.net!howland.reston.ans.net!spool.mu.edu!umn.edu!news
From: "Alan J. Robinson" <robin073@maroon.tc.umn.edu>
Subject: Re: Does AI make philosophy obsolete?  (Was: Quantifying literary progress)
Message-ID: <45393.robin073@maroon.tc.umn.edu>
X-Minuet-Version: Minuet1.0_Beta_18A
Sender: news@news.cis.umn.edu (Usenet News Administration)
Nntp-Posting-Host: dialup-5-115.gw.umn.edu
X-Popmail-Charset: English
Organization: University of Minnesota, Twin Cities
Date: Tue, 29 Aug 1995 14:45:48 GMT
Lines: 120

On Mon, 28 Aug 1995 13:57:46 GMT, 
Bronneberg EM  <embronne@cs.vu.nl > wrote:

>Alan J. Robinson (robin073@maroon.tc.umn.edu) wrote:
>: Kylo Ginsberg  <kylo.fac@mhs.unc.edu > wrote:
>: >jorn@MCS.COM (Jorn Barger) wrote:
>: >>Understanding what computers can do, and how they do it, forces a
>: >>complete revision of all theories of mind and language and even
>: >>ethics.
>: >
>: >... all"), but there is something to agree with here--computers are a
>: >major shift in technology and as such do make us rethink lots of
>: >stuff.  Just as the clock did in its day, just as the printing press
>: >did in its day ....
>: >
>: relationship between computers and the brain is much more than mere 
>: analogy - it's not just a description in terms of the technology at 
>: hand.
>
>If that's so, why don't you tell us what makes this new device so special.
>What IS that little extra the computer has above prevoious analogies. Stop
>just claiming things, provide your readers with some arguments.
>
>Greetings, Miel Bronneberg,
>           embronne@cs.vu.nl
>


Miel:

The relationship between brains and digital computers seems to cause a 
lot of confusion in the AI literature, even at the Nobel Laureate 
level.  Possibly it is because it really is an engineering issue, and 
the majority of scientists don't have an engineering background.  The 
introduction of the Turing machine concept may have also caused some 
confusion, because the purpose of the brain is not to perform 
algorithmic computations, but to function as a programmable 
controller.

A more appropriate model of a brain is a finite state machine:

Input at time t     Input[t]
Output at time t    Output[t]
State at time t     State[t]

Then Output[t] = f1(Input[t], State[t]);
     State[t+1] = f2(Input[t], State[t]);

Norbert Weiner introduced the concept of cybernetics - communication 
and control - in biological organisms just after the Second World 
War, but it never really caught on.  But hierarchical command, 
communication, and control is used in many engineering systems (and 
also many animal and human social systems), e.g. the process control 
systems in oil refineries and chemical plants.  The number of 
levels in process control systems have expanded in recent years, and 
the various levels have their parallels in the brain.

At the lowest level you have the closed loops which maintain 
various variables at fixed values, e.g. pressure and temperature.  In 
animals, the equivalent function is homeostasis, the best example 
being human body temperature held at 37 degrees C.  The next level 
is supervisory control, where setpoints are moved up and down as 
needed.  In animals this would correspond to the elevation of body 
temperature during the fever that is used to fight infection.

At higher levels we are concerned about longer range planning 
issues, e.g. matching supply and demand.  Animals have to 
gather food to meet their energy needs and at the same time evade 
predators, and need to perform various forms of planning.

All these capabilities are based on functional transformation of 
sensory information, which strictly speaking does not involve 
sequential computation.  Nevertheless, the higher one goes in the 
control hierarchy, programmable capability and sequential 
computation in the controller become more and more valuable.  

In fact, evolution solved these problems nearly a billion years ago 
in ways some of which are only just being implemented by control 
engineers.  Single cell organisms have to react to their environment, 
which is often done by the regulation of gene transcription.   This is 
more or less equivalent to the level of intelligence in a thermostat 
or speed governor.  (They're still trying to figure out at MIT whether 
such devices are truly intelligent and are thus conscious <g>.)

In multicelled organisms cellular specialization is possible, and 
the neuron quickly evolved.  Interestingly, neurons perform both 
programmable functional transformations and information transfer from 
one point in the body to another - functions which are separated in 
engineering systems.  Neurons are also very similar to blood cells in 
many respects, and even respond to some neurotransmitters.  Possibly 
this is because there is a lot of communication and control need to 
carry out the immune system's programmed response to infection and 
evolution used some of the smae mechanisms.

The distribution of command and control functions in the brain is very 
similar to that seen in engineering systems, though with some slight 
changes.  The brain's functions are more distributed, possibly to 
allow operation to continue in a degraded manner following 
destruction of part of the system.

A good example is a cat's rage response, which can be evoked by 
electrical stimulation of a particular spot either in the amygdala, 
the hypothalamus, or the periaqueductal grey in the brain stem.  The 
first spot to be discovered was in the hypothlamus, and it was 
erroneously assumed that the hypothalamus was the seat of emotions.  
Even though the hypothalamus is one of the most critical areas of the 
brain, it has also been found that it is possible to slowly 
destroy it over a period of several weeks and have most of its 
functions taken over by other areas of the brain.

Because biological functions have been derived through an evolutionary 
optimization process, there is not the same separation of function 
which is the characteristic of most engineering design by humans.  
(Biological mechanisms are very highly optimized from an energy 
perspective.)  Thus any given biological mechanism may perform several 
functions, which can make it extremely difficult to figure out how it 
works, which applies in spades to the brain.

AJR

