X-Message-Number: 11923
From: Thomas Donaldson <>
Subject: for Mike Perry: the world does not stop for sequential computers
Date: Tue, 8 Jun 1999 23:59:58 +1000 (EST)

Hi Mike!

No, I was not arguing that the neural nets composing our brain were the
same as those we use in computing. They don't have to be, especially for
the kinds of tasks we now use them for. If we want to make an entire
person, that may not be true --- but that is a different question.

As for a neural net being simulated/imitated by a sequential computer,
that's simply something that will not work at all. One ESSENTIAL feature
is that these neural nets respond to the world; since the world will not
wait for a sequential computer to go through all the many operations
needed to simulate a LARGE neural net (I'm not questioning that
possibility for small ones) trying to do that will fail ignominiously.
We're not just doing computing here, we're trying to react to the
world as it comes to us.

You may reasonably ask if a sufficiently large PARALLEL computer might
manage such a task. After all, if it's fast enough we can even consider
making a parallel computer able to predict the weather. My problem
in that case is different: if (as I claim) computers basically deal
with symbols, then with a parallel computer we'll be powerful enough
to work for a while --- until one day our symbols break down when our
parallel computer is faced with something real for which it has no
symbols at all, or even for which its symbols become contradictory.
Remember the black swans! 

			Best and long long life,

				Thomas Donaldson

PS: Incidentally, Canberra is a designed city, with a very nice lake
big enough for sailing in the summer. And they've got not only the
native black swans (which really are black) but immigrant white ones
too. Plus lots of ducks of various species. The lake is surrounded
by a park, as you might guess.

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=11923