Steve, 

 

I am with you here. Remember I always said you and I had things in common? I
have little experience at actually building, or using analog circuits, but
my neuroscientist friend who works on simulating emotions with analogs has
been teaching me for a long time. On my own, I have (finally!!!) defined an
analog implementation of EI. It looks like a neural network, and it can
compute the structures in constant time if   (number of neurons ) >= (number
of elements in the causet). In practice, that means tens of millions, so I
thought of one of those chips with millions of digital microcomputers to
simulate the thing. I don't know enough about analogs to draw conclusions on
my own. In my imagination, my concept could be much better implemented with
analogs than digitals. And perhaps used to simulate the hypocampus. But how
do you do millions with analogs? 

 

Sergio

 

 

 

From: Steve Richfield [mailto:[email protected]] 
Sent: Tuesday, July 10, 2012 1:12 PM
To: AGI
Subject: Re: [agi] Analog Computation

 

OK Ben and Sergio, back to reality.

The simplest hyper-Turing task I know of is simple electric circuit
simulation. With herculean methods it is now possible for supercomputers to
simulate rather simple electric circuits. However, you get more accurate
results by simply building those same circuits with $10 worth of small
parts. Now, simulation is usually used when there is something impractical
about simply building a prototype, e.g. involving massive and expensive
components like substation transformers, is running at such high speed that
it needs to be made into a silicon chip just to test, involves super-high
impedance so that test equipment can't be used, utilizes components that
don't yet exist, etc., etc.

Note that I was involved in developing the first program to solve the
Schrodinger equation, working with Dr. Ira Karp at the University of
Washington Physics Department. Yes, like in electric circuit simulation, it
IS possible for Turing computers to eventually simulate hyper-Turing
phenomena, but the simulation is imperfect and ever SO slow.

Steve
===============

On Tue, Jul 10, 2012 at 7:06 AM, Ben Goertzel <[email protected]> wrote:


This is ground that has been well-trodden in the recent literature on
hypercomputation.

Mathematically, yes, analog computers (and analog neural nets) can in
principle do hyper-Turing computation

Quantum physics appears to rule out this kind of analog computing existing
in reality, but general relativity would permit it... and as you know these
theories have not yet been unified.  String theory and loop quantum gravity
would also, according to my best understanding, not permit it...

Since the totality of scientific data consists of a large finite set of
finite-precision numbers, there is no possible way for science (as presently
conceived) to validate or refute the hypothesis that hyper-Turing computers
(of the analog or other variety) exist physically.  In this sense, the
hypothesis of trans-Turing computing in the brain or any other physical
system is non-scientific.

-- Ben G

 

On Tue, Jul 10, 2012 at 8:57 PM, Sergio Pissanetzky <[email protected]>
wrote:

Ben, 

 

You asked recently for comments on analog computation (I forgot what thread
that was, so I started a new one). My comment is on the need for a
comparison between Turing computation and analog computation. Of course, AC
can do all that TC can do, but can it do more? I believe it can. TC is
particularly weak in everything related to binding, associations, and the
resulting structures (BAS). Anywhere that BAS are needed, humans are called
for help. Have you noticed? It never fails. I have compiled a list of
problems that are "easy for humans but very difficult for computers to
solve", the GUAPS, great unsolved automation problems of software
engineering. They all critically depend on BAS. They include OO analysis and
design, object recognition, self-programming, and many others. Turing
himself was concerned about this limitation, and he wrote extensively about
it in the context of morphogenesis but never solved the problem. 

 

The GUAPs, of course, include the invariant structures that our brains make
(I know a chair is a chair even if it is upside down => invariance under
transformations).

 

It seems to me that AC can do more than TC. Because physical systems do
self-organize and make structures. Physics even has theorems about this. If
so, AGI would be squarely outside the domain of TC, and strictly within that
part of the domain of AC that is not TC.

 

Sergio

 


AGI |  <https://www.listbox.com/member/archive/303/=now> Archives
<https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> | Modify
Your Subscription

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57>  


 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> AGI |
Archives | Modify Your Subscription

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 



 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> -- 
Ben Goertzel, PhD
http://goertzel.org

"My humanity is a constant self-overcoming" -- Friedrich Nietzsche


 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> AGI |
Archives | Modify Your Subscription

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 


-- 
Full employment can be had with the stoke of a pen. Simply institute a six
hour workday. That will easily create enough new jobs to bring back full
employment.

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57>  


 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> AGI |
Archives | Modify Your Subscription

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57> 

 <https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57>  




-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to