From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!wupost!uunet!idtg!dow Thu Dec 26 23:56:58 EST 1991
Article 2259 of comp.ai.philosophy:
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!news-server.csri.toronto.edu!rpi!usc!sdd.hp.com!wupost!uunet!idtg!dow
>From: dow@idtg.UUCP (Keith Dow)
Newsgroups: comp.ai.philosophy
Subject: Re: Scaled up slug brains
Message-ID: <361@idtg.UUCP>
Date: 18 Dec 91 23:11:40 GMT
References: <351@idtg.UUCP> <60637@netnews.upenn.edu> <359@idtg.UUCP>
Organization: Integrated Device Technology, Santa Clara
Lines: 81

>Finite cellular automata can exhibit phase transitions.

Finite cellular automata are artificial, so you can say anything you want.
  
>I've yet to see a statistical mechanics textbook make the
>claim that there are really infinitely many molecules in their ideal
>gases.

The distribution functions assume an infinite system. 

>I've posted the same reference on pumped phonons condensations three
>times now. 

That was experimental paper?  I thought it was some person talking about
how they think the human brain is.  So somebody has put microprobes inside
the human brain and said "Ah! A phonon pump!".  I don't think so.

>>Froehlich is not a chemist.  He's a physicist.  He's not a biologist,
>>though, which is why others do the relevant experiments.
 
Which biologist published the paper saying here is a phonon pump?
Here is how I measured it.  Here is the data I recorded.  Since phonons
aren't charged, they won't show up using electrodes.  The paper should
be very interesting. 
 
>his seminal work in non-equilibrium thermodynamics?
 
I guess that is why I put in the phrase "in or NEAR thermal equilibrium."

>You contradict yourself.  Negative temperatures show up when talking
>about finite systems, not the n->oo that you say is a real infinity.

Negative temperatures show up wherever you want to use them.  It is
just a dirty trick used to do calculations.

>>Bose-Einstein condensation refers to bosons occupying the lowest
>>energy state.

>What I said.  It does not refer to how they got there, whether by phase
>transition or pair formation or pixie intervention or too much keyboard
>drool.  Either the bosons are all in the ground state, or they aren't.
 
Since there is no limit to how large a phonon wavelength can be, the phonons
have no ground state.  Therefore Bose-Einstein condensation doesn't apply 
to them at all.

>>Fine, then we all agree that the brain is a machine which will soon
>>be understood.
 
>The question has been *what* kind of machine.  "Soon?"  AI folks have
>been singing that tune for three decades now.  Soon on geologic time
>scales, for sure.


I would not trust the classic AI folk farther than you can throw their copy of
Perceptrons by Minsky and Papert, with Minksy and Papert tied to the book.
The neural network people hold the future.
 
Unfortunately computer science classes for the most part only teach how
to reinvent the wheel.  That is why the software industry does not have much
of a component industry.  Also, that is why the software industry has moved 
forward at a slugs pace.
 
On the other hand, hardware has advanced much faster than software.  Since
neural networks are programs which program themselves, or hardware which
programs itself, the bottleneck of the programmer is minimized.  So I see
neural networks becoming a dominate force in the next twenty years.  Of 
course classic AI will improve a lot with neural networks.  Eventually
we will see both working together in a semi-seamless enviorment.

Sixty years ago is 1931.  Vacuum tubes are the state of the art.  A guy
in the midwest has yet to build the first electronic calculator.

Twenty years ago the first microprocessor was built, the 4004.

Fifteen years ago, the Apple-1 was first built.

Nothing interesting happened five years ago.

Considering our progress in the past sixty years, I think my estimate
of sixty years from now to understand the human brain is good.


