Alt-BEAM Archive

Message #04223



To: beam@sgiblab.sgi.com
From: Tom Edwards edwards9@hotmail.com
Date: Thu, 03 Jun 1999 12:50:33 PDT
Subject: [alt-beam] Re: Transistor MicroCore



>From: James Taylor
> Finished going over Tilden's living machine paper
>for about the 5th time. Still haven't discovered the
>secrets to his Spider.

Check his patent (you can find links to various copies from the Solarbotics
site) 'cuz I think it adequately describes Spyder. There may be
additions/changes to the device but the patent covers the basic control
circuitry.

>Anyway, he mentions neurons made of transistors.

If you were to design Nv and Nu circuitry directly into an integrated
circuit, it only takes two CMOS-style transistors to implement the "neurons"
(see Suzanne Still's papers). Under those circumstances, you could perform
the function of a hexcore using only twelve transistors. Practically
speaking, however, a 74xx14 chip has something more like 54 transistors so
until there are BEAM chips this "simplicity" has been exaggerated by a
factor of about 4.

There have been previously discussions about the lack of merit in building
nervous net circuitry using discrete components but it is nonetheless
possible.

Wouter wrote:

>I don't know what Mark Tilden did in the early years of Beam,
>maybe he used transistors, maybe not; maybe he used them
>sometimes to see what the differences would be.

To do the best of my knowledge, he went directly to 74xx14's -- he
explicitly mentions "Schmitt inverters" in the patent which was granted in
1994 (five years ago -- ought to have been early enough in the BEAM
universe.)


Tom


_______________________________________________________________
Get Free Email and Do More On The Web. Visit http://www.msn.com

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com - Simplifying group communications



Home