Alt-BEAM Archive
Message #10875
To: beam@sgiblab.sgi.com
From: BUDSCOTT@aol.com
Date: Fri, 25 Feb 2000 18:28:25 EST
Subject: [alt-beam] The Almost Complete Walker
As you might have been following, i have been making a walker with the
almost complete walker circuit. I got rid of those infernal MPJA worm gear
motors and got some really really really nice display motors to do the job.
the only problem is that the really really really nice display motors have a
really really really high gear ratio. the walker does do anything except
slightly wiggle. how can i lengthen the timing of the bicore to compensate?
Thanks a ton!!!
-"Sad cause my #$%&!! walker won't work in Illinois" or just Spencer
10876 Fri, 25 Feb 2000 17:31:26 -0600 [alt-beam] Re: BEAM: Tendency toward miniaturization beam@sgiblab.sgi.com Terry Newton At 12:22 PM 2/25/00 -0700, Robert Morris wrote:
> ...
>And I agree that using digital logic for large neural nets will only
>produce more digital logic,
>not a living brain. But if you reduce the size and power requirements and
>use analog circuits
>for the neurons, the possibility for life is much higher.
Or at least take an analog philosophy, digital logic / uP can accomplish
it to a degree if set up right but there's something really sexy about
doing it in analog. Bigger phase space but got to Keep It Simple
or Can't Do It, which usually drives me to code.
>I believe that analog circuitry(at least partially) and a robot body is a
>requirement for a
>machine to "become" alive or have a consciousness. This is because a
>digital computer sitting
>on a desk cannot gain any experience or be influenced by the constantly
>changing analog reality
>that we live in.
I'd argue about that last one but does not pertain...
>This is why I have been interested in BEAM, while not perfect, Nv/Nu
>circuits can be affected
>by the outside environment to a certain extent. What is needed is for
>someone to develop
>a VLSI chip containing large quantities of Nv/Nu circuits in which the RC
>time constants
>can be electronically adjusted and the interconnection between them can be
>changed and
>re changed to any configuration desired. Then I believe using a number of
>these chips, it
>may be possible to see some very lifelike qualities emerging.
A hardware version of a Self-Wiring Bicore Array... I posted some
results at: http://www.nc5.infi.net/~wtnewton/otherwld/selfwire.html
To have it in real analog would rock! But probably should have a
digital interface for configuring the connections, evol algs and
things like that are fairly easy to code but would be difficult to
hardwire. On the other hand the neurons themselves are a difficult
thing to finely represent in code so the combination would be
most useful.
I got good results learning simple but varied obstacle avoidance
responses using the crude equivalent of 8 Nv units connected up
as 4 bicores. It appeared to learn the positions of objects and
avoid them but it was a clever trick, finds a sequence that
works and sticks with it. Whether it will scale to produce
meaningful results with larger networks, I don't know.
>To understand a little why I believe a-life is possible, here is a thought
>experiment:
> [snip]
I can't better Phillip's response.. but I do believe a-life, at least
on some degree, is possible. Some might argue it already exists.. who
am I to say that when I evolve corewarriors or robot brains that
something special is not occuring, I cannot tell for sure, but
special or not it sure is fun!
Perhaps we get lost in definitions and it makes for debatable points
when we're really agreeing, Jim's treatment of thought objects was
nice, and we do need to dream a bit to accomplish things. And yes
perhaps I'm a bit westernish in my convictions (working on it...) but
I have -not- declared that artificial consciousness is not possible,
just that we'd be wasting our time trying to duplicate -human- level
consciousness. Or even dog-level. Just a little research into the
complexity of nature shows how futile that would be. But awareness
isn't black and white, it can exist in infinite degrees. The robotic
greats did not and are not wasting their time for they declared
only an amount of awareness suitable for robotic implementation,
and for the most part achieved it.
One thing that is worthy to note on the subject of robot soul...
I noticed that for a given environment, body and set of goals,
the results are often the same even if implemented differently.
For example, both the bicore array and conventional RL learn
to go around in circles if placed in a confined area. When a
robot is given a way to modify itself and the rules can be
satisfied, the system tends to soak in aspects of its
environment that are not readily apparent. And remembers.
I don't know exactly what it is, but it probably has just
a little more right to exist than the average hookup.
Then there's the hidden interactions that take place when
nervous nets are allowed to influence each other, even when
wired fixed there's still something unique going on in there
that I can't quite put my finger on, it represents knowledge
in a circulating pulse and when the pulse is hidden and read
out in a way that doesn't reveal its position I suspect
strange things are allowed to happen.
>P.S. The quantum experiment I talked about is real, I'll see if I can
>find the article if you
>would like.
I'd appreciate it, I'm into that kind of stuff :)
Terry Newton
Home