X-Message-Number: 8638
Date: Mon, 29 Sep 1997 07:41:27 -0700
From: Peter Merel <>
Subject: Analog Neurons? And whoops again.

I wrote,

>This is obviously similar to the general halting problem - ie. we can
>enumerate the programs that halt, but not the ones that don't halt. 

Wow, my old CS lecturers must be whirling in their staff rooms. 
No, programs that halt are not RE. Sad waste of a good education :-(

Thomas Donaldson writes,

>It turns out that a neural net with real-number weights on 
>each connection cannot be imitated by a Turing machine. 

Thanks for digging that out Thomas. Verrry interesting ...

>The implication that our brain's neural nets might do some things 
>impossible for a Turing machine bears a lot on the various controversies 
>that have been going on in Cryonet.

... but impractical. This work does not imply that the neurons in your noggin
have the ability to work with infinite-precision reals. Analog computers don't
actually exist - just limited-precision approximations to them. Hmm. Well, 
perhaps the universe itself, if it is unbounded in spacetime, may be an 
analog computer, but it seems unlikely that neurons could take advantage of
this - no?

Peter Merel.

Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=8638