Re: Random thoughts on MEMs based robotics

Is anyone seeing this? Links to some of my posts have been broken..
just wondering.

Anyway, someone asked of this technology can it be smart enough to do
something interesting? That is, can you put a program in a device
that consists of 27 billion discrete elements? (this device being the
27,000 'cell' with each cell containing 1 million components, 'dust
mite' sized robot.)

I think so.

To prove this requires advanced math - which is beyond the scope of
usenet to teach - haha - and maybe impossible - but I learned from a
great series of books that fell into my possession back in the 60s.


There was a great series of books published back in the 1940s - Modern
Algebra - Ungar Press - Edited by VanderWaerden. I picked up the
whole set at a garage sale from 1940.

Garage sale's are interesting things when they take place in Princeton
NJ around the campus... Sort of like what a garage sale would be like
in Eureka I suppose.

Anyway the old fart who sold them to me pointed out that they were
translated from the original German by the US Naval Academy, and that
you won't find anything like it these days. Why? I asked
incredulously. Well, this was published and distributed in the public
interest by authority of the Attorney General of the United States
under License No. A-1475 - see? It says so right here. What's that
mean old-timer? lol. Well, the US government circulated these to
find out who might respond to them, and then they took the best and
put them to work in various capacities. What capacities? Secret
capacities boy.

Then, they either classified the stuff or took other measures. The
math they couldn't classify. It said so in the law. But what they
COULD do, is keep the copyrights once they had them, and not publish
any more books like this EVER again.

Why would they do that? The old man laughed as he shifted his weight
in his wheel chair. Same reason they introduced the NEW MATH to the
schools about that time! The only way they could keep every Tom ***
or Harry from building an atom bomb was to dumb us all down boy.

I regarded the old timer. How much you want for these books? This
story isn't increasing their value you know, I'm not THAT dumbed
down. Hell I'll GIVE you the books if your promise one thing. What's
that? STUDY THEM! Deal! And that's what I did. haha..

So, anyway, theres a concept that's very interesting and useful - its
called CONFIGURATION SPACE. And this can apply to an assemblage of
atoms and neutrons - and tell you how to creat interesting devices

But the math can also apply to a machine system of repeatable
elements.

The sensors the machine has can be thought of as a single entity, that
has a number of configurations possible.

The outputs of the machine can also be looked upon a single vector,
that also has a number of configurations.

Its also easy to see that sets of sensor configurations, and sets of
responses, can also be a larger universe of configurations in the
appropriate configuration space by grouping instantaneous responses
together. That is the SEQUENCE a,b,c is different than the SEQUENCE
c,b,a or b,a,c etc.. And these sequences or groups can form a
larger configuration space.

Well, ANY function that connects a sensor configuration to an actuator
configuration can be considered a controlling program for that
automaton - or device. See? We can imagine that under this
condition - which gives rise to this sensor configuration, we want the
machine to respond this way - which gives rise to this actuator
configuratoin. So, all we're doing is making a patch connecting THIS
sensor configuration to THAT actuator configuration.

Now, any configuration can be written as a number - properly
constructed.

Well, the simplest way to write such a program is to make a list of
numbers that represent the machine's responses - and put them at
locations in the list where their address, or index, is equal to the
sensor configuration associated with that response.

That is, if you want configuration 7 on the actuator array whenever
the sensor array reads 3 - then all you have to do is write the number
7 in record 3 - and arrange the sensor array value to index to that
position on the list. Do this for all possible configurations - and
you have a complete program.

Now this is the least compact way to do this. Once you do it, you can
apply hypergeometric techniques to 'collapse' the HUGE list into
something manageable. In fact, you can use simple rules to CONSTRUCT
the list given GOAL DIRECTED BEHAVIORS.. and to allow the machine to
LEARN from its environment so that it can more efficiently achieve the
goals it is given.

But that's even more beyond what I wanted to say.

The point is, a rather simple mechanism that reads and writes lists of
numbers, can serve to implement a program of arbitrary complexity!
VonNeuman machine someone might insert here - but vonNeuman's machine
is only one of a class of machines. Other machine classes are
possible.

In any event, the interesting thing here is that the DNA/RNA codon
trick to store terabits of information in a single cell - this trick
can be used to store terabits of CODE in a single cell - which could
lead to quite sophisticated behavior!

The human brain is thought to store information by dint of the
interconnections neurons make with other neurons via dendritic and
synaptic connections - and the amount of information processed by the
brain may be on the order of several hundred thousand terabits.

This is a different structure than a list processor described above.
But only a few thousand of the 27,000 cells would be able to process
information on the scale of a human brain using it!!

That is, EACH of these tiny mites, properly programmed, would be as
complex in their behavior as a human being!! (potentially)

Just something to think about.

If the mites were programmed, or programmed themselves, for self-
preservation - being a higher priority than service to their creators
- they might be very difficult to deal with.

Especially in populations of quadrillions of creatures...




.

Quantcast