Alt-BEAM Archive
Message #11445
To: beam@sgiblab.sgi.com
From: "Thomas Pilgaard" ascii@hum.auc.dk
Date: Tue, 7 Mar 2000 19:41:05 +0100
Subject: [alt-beam] Re: Self-Awareness/Conciousness
Although Bob and Jean already has pounded at it :) I'd like to put in my two
cents aswell.
I seem to remember a definition of self-awareness that sounded something
like "If X knows P then X knows that X knows that X knows P".
As I've said, Bob and Jean have been at it so I'll try for once to control
myself and leave it with that. Hnnngh-argh, no, I gotta put in my two cents.
Exactly !what! does a beam robot !know!? IMHO it doesn't as such know
anything; it reacts specifically to circumstances and the idea of it being
aware of anything is far fetched. Therefore - though I find the idea
extremely interesting - the "inner loop" does not know of or is aware of
anything. It receives input from something to which it reacts. It has no
sence of where the input comes from and what to make of it.
Now the application of nested nervous nets is a totally different ballgame.
Cheers,
Thomas
> You know i read in that Smithsonian Article about BEAM something
> just blew
> my mind off. Brossl Hasslacher (Tilden's collegue in BEAM)
> proposed the idea
> that you can take nervous nets to a new level. In theory, by
> implementing a
> nervous net WITHIN another nervous net, you actually have a robot that is
> aware of its own existance. See, the outer net is directly
> interacting with
> the physical world, while the one within is only in contact with
> the outer.
> This means that the inner net is AWARE of the outer one, which is
> in turn in
> the PHYSICAL world. Thus the robot is AWARE OF ITS PHYSICAL EXISTANCE!!
>
> Kind of sends shivers down my spine.....
>
>
> www.geocities.com/beambotix
>
> ______________________________________________________
>
11446 Tue, 07 Mar 2000 13:35:48 -0500 [alt-beam] Re: Quick-Change components beam@sgiblab.sgi.com "Sathe Dilip"
"Feser, Jason" wrote:
>
> Yup. They are pin sockets. I've managed to melt about 1000 of them out of
> old pc hardware with a pencil torch, but you can buy em' if you want. Dave
> sells them over at http://www.solarbotics.com.
I have a pencil torch but never thought of this use. Have you tried
salvaging ICs, transistors etc. What is the mortality due to heat if
one is quick enough? I have read on another newsgroup about someone
using a barbecue to remove components enmasse. This person claimed not
to lose that many due to heat.
Dilip
----------------------------------------------------------------------
The early bird may get the worm, but the second mouse gets the cheese.
----------------------------------------------------------------------
11447 Tue, 07 Mar 2000 12:37:11 -0600 [alt-beam] Re: MicroCore vs. BiCore beam@sgiblab.sgi.com Richard Piotter A Microcore is a loop of Nv neurons. It can be a loop of 4, 6, 8, or
more, and also can be a 2 Nv loop, but since bi means 2, it's called a
Bicore. I've heard the word hexcore and octocore used, but not often.
Most people (myself included call them, 6 Nv and 8 Nv (and up)
microcores). Bicores are popular cause they are easily joined together
so one effects another. Like stacking them. Microcores of 4 Nvs and
larger tend to be used in walkers and sometimes chains or bicores are
attached to it. bicores are the same. the difference is the number of neurons.
There are also differences in the chips. the 240 can drive low current
motors, and higher current ones when gates are tied. the 74HC14 needs
drivers and has fewer gates. The 240 can work as a "Suspended" bicore
too. I'm not going to try to exp0lain much about it, but basicly, you
tie the resitor points together, usualy using a pair of back to back
photodiodes. They form a voltage divider.
"Scolman,Jim" wrote:
>
> Hi, well I am pretty confused at this point, I am new to BEAM and have
> assembled two kits to this point. Now I want to build a way cool
> walker.......I am thinking solar, with two servo motors, and using the
> 74AC240 chip so I don't have to use a motor driver.......I want to use the
> big solar cell and the motors from Solarbotics......my new confusion is
> Bicore or MicroCore........I had thought a MicroCore was a chip with several
> Bicores "on" it?.......so which is which and what's what ?.......I HAVE been
> reading the FACs and tutorials and collecting the diagrams, but I am still a
> little fuzzy. Thanks for your help........JWS.
>
> Jim Scolman
> AMS Services, Inc.
> Bothell, WA
> 425-402-1000 x353
> 206-910-6095
> scolmaji@amsworld.com
> Network Administrator
>
> -----Original Message-----
> From: Wilf Rigter [mailto:Wilf.Rigter@powertech.bc.ca]
> Sent: Tuesday, March 07, 2000 7:26 AM
> To: 'beam@sgiblab.sgi.com'
> Subject: RE: MicroCore vs. BiCore
>
> If damn PNCs are you only concern, you can add 2 diodes to any microcore to
> make it PNCless.
>
> The (recycled) PNCFREE microcore starts instantly and is always on alert to
> automatically neutralize saturation.
>
> Also note the simplified LED circuit: Since only one LED is on at one time,
> the 4 LEDs can share one resistor.
>
> wilf
>
> <>
> > -----Original Message-----
> > From: David Perry [SMTP:davidperry@geocities.com]
> > Sent: Monday, March 06, 2000 9:01 PM
> > To: beam@sgiblab.sgi.com
> > Subject: Re: MicroCore vs. BiCore
> >
> > my favourite reason for the bicore is you don't need no damn PNC! Sure its
> > fun to play with a microcore, but it always mess up the circuit when i
> > bump
> > the resistors.
> >
> > David
> >
--
Richard Piotter The Richfiles Robotics & TI web page:
richfile@rconnect.com http://richfiles.calc.org
-- Make Money by Simply Surfing the Net or responding to E-Mail!!!
-- Click below!!!
http://www.alladvantage.com/go.asp?refid=ATL147
http://www.spedia.net/cgi-bin/dir/tz.cgi?run=show_svc&fl=8&vid=329630
11448 Tue, 07 Mar 2000 11:43:32 -0700 [alt-beam] Re: nv's and nu's beam@sgiblab.sgi.com "John A. deVries II" At 12:13 AM 3/7/00 , Bruce wrote:
>Steven Dang wrote:
> > I'm trying to figure out the difference in the schematics for a
> > neuron in a nervous net and a neuron in a neural net...
>Hi, Steve. Just happen to have (see attached) ...
And good diagrams they are. They are an excellent answer to Steve's question.
-------------------------------------------
However, the question in the body of Steven's message is ill-formed and
doesn't correspond to what he had in his >Subject line. "Neural nets" are
TRULY different beasts than BEAM-style Nv/Nu nets (particularly the latter
sort).
That Tilden has chosen to use the nomenclature of the neural net community
is sadly unfortunate -- it leads to quite a bit of confusion. For example,
"proper" (ok, conventional) neural nets can be trained by altering the
strengths of connections between neurons; neural nets are very generally
(and boy do I mean generally) specified by three layers of neurons that are
arbitrarily connected between any two layers; neural nets don't have the
integrative/differentiating nature that Nv/Nu nets have as a result of the
capacitor included in the latter's circuit.
There are exceedingly few Nv/Nu circuits that show anything beyond a ring
or master-slave (or phase-driven) topology and in either case they act as
complex oscillators: I believe that it would be difficult to use them for
building an "expert system" which is what neural net research eventually
lead to.
Descriptions of some of the "proper" neural net-type neurons can be found
on pages such as http://burks.bton.ac.uk/burks/foldoc/29/67.htm
(McCulloch-Pitts neuron), http://burks.bton.ac.uk/burks/foldoc/36/50.htm
(Hopfield networks), ftp://ftp.sas.com/pub/neural/FAQ.html#A2 (Subject:
What is a neural network (NN)?) and the like. The web is full of
references like these; a particular classic is the 1986 book edited by
Rumelhart and McClelland, "Parallel Distributed Processing : Explorations
in the Microstructure of Cognition : Foundations" which can be used as a
doorstop if you happen to get bored of cognitive research .
Another word for "conventional" neural nets is that they are
"connectionist". Mr. Tilden has declared that some of his devices are
"connectionist" in nature -- this is true enough given that it is his
_fingers_ that are determining the new connections (and that one can use
wires and resistors to connect the elements of the circuit in different
ways, providing different behaviors). In an "ideal" conventional neural
net the connections are determined automagically, much like the
electrically activated arms in Ashby's homeostat. Chances are that the
argument will be made that an equivalent sort of "state memory" is stored
in the phase relationships of an Nv/Nu circuit (this assertion is yet again
stated in the Smithsonian article) but I haven't seen any convincing proof
of that. Furthermore, there is nothing whatsoever in any of the Nv/Nu
circuits that correspond to the variable (and, in fact, trainable)
connection weighting that is an essential part of "conventional" neural
nets and hardly any circuits that exhibit the multiple inputs that can be
made to be made to a given neural net neuron.
Just being your old cantankerous Heretic,
Z
11449 Tue, 07 Mar 2000 13:45:21 -0500 [alt-beam] Re: Quick-Change components beam@sgiblab.sgi.com "Sathe Dilip" Digi-key has several pin receptacles (diff. dia) available from Mill-Max
(page 124 of Jan-March catalog).
I have used socket pins removed from D connectors (from old PC cards)
for use with thicker leads.
Dilip
Bruce Robinson wrote:
SNIP
> I've experimented with these a bit and I discovered that they work best
> with devices that have square or rectangular pins (e.g., IC chips). When
> you're dealing with round wire leads, the diameter of the wire will
> determine how well the socket grips.
>
> I find many of my resistors get vibrated loose from machine tool
> sockets, so I've stolen one of Chiu's ideas ... use a conventional IC
> socket and bend the resistor leads around to suit the socket spacing.
> Some (but not all) IC sockets will give a secure grip to many different
> wire diameters.
----------------------------------------------------------------------
The early bird may get the worm, but the second mouse gets the cheese.
----------------------------------------------------------------------
11450 Tue, 7 Mar 2000 11:01:16 -0800 Re: Quick-Change components beam@sgiblab.sgi.com Sathe Dilip [mailto:sathe_dilip@bah.com]
"Feser, Jason" wrote:
>
> Yup. They are pin sockets. I've managed to melt about 1000 of them out
of
> old pc hardware with a pencil torch, but you can buy em' if you want.
Dave
> sells them over at http://www.solarbotics.com.
I have a pencil torch but never thought of this use. Have you tried
salvaging ICs, transistors etc. What is the mortality due to heat if
one is quick enough? I have read on another newsgroup about someone
using a barbecue to remove components enmasse. This person claimed not
to lose that many due to heat.
Dilip
----------------------------------------------------------------------
The early bird may get the worm, but the second mouse gets the cheese.
----------------------------------------------------------------------
11451 Tue, 7 Mar 2000 14:38:01 EST [alt-beam] Re: Was lobster, now bees beam@sgiblab.sgi.com Bumper314@aol.com In a message dated 3/7/00 10:52:37 AM Mountain Standard Time,
Bruce_Robinson@telus.net writes:
> This is certainly in reach of the BEAM community. Richard Caudle's HPV
> idea can be adapted to this purpose. Sound is the communications medium
> ... the robot that finds food sends out a periodic "chirp" and the
> robots that hear the chirp turn towards the sound.
>
> The circuits to detect and create the sounds exist (thanks to Wilf). All
> that's needed is to define "find food" (in robot terms) and adapt the
> sound circuits.
Thats a great idea. Sounds would be the best way to communicate between the
bots. And since we are all talking about this together I think if we get a
standard plan out then we could have others contribute to the whole RJP. I
believe and idea about this was brought up earlier, but didnt go
anywhere...we'll see about this one
Steve
11452 Tue, 07 Mar 2000 14:37:14 -0500 [alt-beam] Re: Transistor codings beam@sgiblab.sgi.com Marcus Cole At 14:19 07/03/00 GMT, you wrote:
>whats written on them????
>
>
I have over 30 of them, so i'll only list a few of them.
-C1675
K13IC
-C1674
K 3ZC
-STS
8050
C 02
-C9018
H 5I
-A733
P241C
-C945
P13ZC
-N5016
R 21
-K 2F
C9012
H
Also, is there any books or web page(s) database containing all the codes
for any parts i could find? That would be helpfull to anyone who wants to
recycle old electronics gizmos. Thanks in advance,
Marcus Cole
11453 Tue, 7 Mar 2000 14:49:10 EST [alt-beam] Re: nv's and nu's beam@sgiblab.sgi.com JVernonM@aol.com In a message dated 3/7/00 1:55:48 PM Eastern Standard Time, zozzles@lanl.gov
writes:
> That Tilden has chosen to use the nomenclature of the neural net community
> is sadly unfortunate -- it leads to quite a bit of confusion.
So, what else is new :)?
See ya,
Jim
http://www.geocities.com/SoHo/Exhibit/8281/beamart.html
ICQ# 55657870
11454 Tue, 7 Mar 2000 12:08:24 -0500 Re: photoresistors Bruce Robinson
> William Cox wrote:
> >
> > Does any body have a schematic for a phototropic head
> > or ant that uses photoresistors instead of diodes?
>
> Here's one that has worked well for me. The photoresistors are the ones
> to the right of the 1 meg resistors.
>
> You'll need a slow oscillator of some sort to generate pulses for it ...
> easy enough to make one from a Schmitt inverter.
>
> In operation:
>
> - a pulse causes two parallel Nv's to go active simultaneously.
> Since the two outputs feed two sides of the head motor, the
> head won't move.
>
> - the photoresistors have a much lower resistance than the 1 meg
> resistors, so they shorten the usual Nv delays considerably. With
> different amounts of light striking the photoresistors, the Nv on
> the brighter side will time out more quickly and go inactive.
>
> - this will immediately start the head moving, until the slower Nv
> times out.
>
> - In order to make the motor run longer, the photoresistors are
> connected via diodes to the opposing Nv outputs. This means
> when the faster Nv times out, it effectively takes the
> photoresistor out of the slower Nv. This slows the Nv right
> down.
>
> - Each pulse from the oscillator moves the head a little.
>
> - The effect is to move the head in short bursts until the
> resistance in the photoresistors is balanced.
>
> Point the photoresistors slightly away from each other (i.e., not
> parallel) and shield them from overhead light. You will also want to
> have well matched phtotresistors. The circuit responds very nicely, but
> is highly sensitive to anything that unbalances the circuit.
>
> Note that you can't feed a common H-bridge with this circuit, as it will
> turn on both sides at once & fry the bridge. With a low current motor,
> you could use a 74HC245 to drive the head; or try one of Wilf's
> "smokeless" bridges on the Beam Heretics site (Rigter Archive). I
> believe Wilf also created a similar circuit to this one which used a
> single chip to provide some limited head driving capability.
>
> Enjoy,
> Bruce
Home