From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sun Oct 07 2001 - 22:23:20 MDT
I do not believe there has been any "proof" that a quantum computer
can function as a Turing machine and can therefore function as
any computer.
There are certain computations, such as factoring, that have been
mapped into quantum algorithms which can therefore be solved using
a quantum computer. So a quantum computer, to the extent that I
understand them, can only be used to solve a limited subset of
the phase space of what is computable.
So I do not believe a comparison between the brain as a neural
network and a quantum computer is a useful comparison. As I've
discussed in my Matrioshka Brain papers, you are going to get
a factor of 10^5 increase by going to reversible rod-logic, which
combined with a volume reduction of 10^3 gets you at least 10^8
improvement in throughput (probably more if you allow for the fact that
you can move from ion-current signal propagation delays to speed-of-sound
in diamond for nearby processors to speed-of-light for distant
processors). Going to helical-logic gets you a factor of >10^11
and if you can get ballistic heat removal (which is starting to
look possible given recent buckytube simulations), another factor
of 10^9 or so.
So in terms of throughput, I'd guess the limits are > 10^20 increase
without having to move us to qubits or sub-atomic engineering.
Of course discussions regarding throughput per unit of real time
get rather valueless once one has uploaded since you are then
living until the heat death of the universe.
Now of course, a nano-brain with a qubit add-on subprocessor
would of course be useful, particularly if your job is gathering
intelligence information. :-;
One thing that remains to be determined is the degree to which
the brain is "packing bits". It looks like the neurons may
be moderately sensitive to the timing of the firing of neuronal
pulses -- if so, there is more going on than you can account
for by simply counting neurons. Also important is the fact
that the brain is *not* primarily a computation engine --
its much more of a communication engine. Each of the 40
billion neurons can have inputs & outputs ranging from 10^3-10^5.
And there are rumblings in the neuroscience community that the
glial cells are doing more than providing "nutritional" support.
Unraveling the whole system is going to be a nightmare and I'm
very glad that people like Anders and Eliezer are the ones
wrestling with it and not me.
Robert
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:11:15 MST