Re: Best Programming language for AI in Nanobots


In article <11tqgfosp5kt11f@xxxxxxxxxxxxxxxxxx>, rhooker123@xxxxxxxxxxx
says...

> As for flaws in my analogy, I lean to hard singularism, though I would
> never give money to it, and I see evolution as a process of information
> optimisation. The analogy between coding with a evolution process and
> sending some living things in a space where they can fight it out is
> not very different, you have information (either code or DNA) and you
> have goals and the one's who meet the goals best win.

In practice, it is extremely different.

In the physical world, biological evolution is biological evolution. It
follows the laws of physics and favors those strategies which allow it
to reproduce.

In the software world, genetic algorithms and genetic programs can
become much more subtle. They take inspiration from physical biological
evolution, but are not constrained to it. The metaphorical environments
can be manipulated in ways that the real world cannot, as can the
representation of the various biologically inspired objects and
operations.

And again I note that if your view is so broad as to make all
optimization processes equivalent to "evolution" then you are removing
great swaths of numerical optimization techniques from the basket of
acceptable techniques. Why is a genetic algorithm out, while simulated
annealing, gradient minimax, or ant colony optimization in? Where is
the dividing line?

> I build massive systems to store the TBs produced by entire businesses
> or government agencies and my view on chips is rather high level, so I
> am more than happy to defere to you and take you statements seriously.
>
> The main problem I see is NOT engineering though, its insurance. How
> can you get a company to underwrite the production of a nanobot which
> you can't produce a full working spec for?

WHat is it about this process that you believe impedes the process of
writing full working specifications?

> Its one thing with
> circuits, since a circuit can't reproduce itself and errors will be
> local. As for testing experience from drug tests show these can never
> find all possible errors. How many times does a drug that was
> perscribed by doctors turn out to be unsafe.

Again, I ask:

o What about non-replicating designs, built by something else?
o What about nanobot subsystem design?

As for the task of finding all possible errors, testing experience from
Intel on rational designs (at least as far as I know) shows that you
can't find all possible errors, period. It wasn't that long ago that
Intel issued a great big corporate, "Oops!" when they detected a subtle
floating point error in their chip.

Well, I guess it depends on how you define "that long ago" since I seem
to remember being in grad school at the time, which would put it about
ten years ago, now. The point remains the same, though-- the complexity
of design is the problem, not the method of design. This problem gets
worse and worse all the time. It may also surprise you to know that
genetic techniques are applied to testing, as well. I have on my shelf
a book about Genetic Algorithm in VLSI design with at least one chapter
devoted to GA assisted testing. (Can't provide details-- haven't had a
chance to do more than skim it.)

> Given the GOO problem I think the industry is going to demand
> extraordinary safety checks on nanobot programming.

As well they should!!

> Solutions can be
> evolved over time via CAD but those solutions will need to be fully
> understood and proven that they won't run off in to some crazy GOO
> madness, unless there is some proven self destruction module that gives
> the little thing a 5 hour life or something.

I find the premise of accidental but faithful self-replication where
none was intended to be a little far-fetched. If that were easy, then
the creation of life in the lab would be trivial, not exceptionally
difficult and unachieved as yet. Not to mention, it wouldn't be a mark
of skill to sit down in some arbitrary programming language and write a
quine.

(A quine is a piece of software which generates, as it's output, a
perfect copy of itself without resorting to cheap tricks like reading a
copy of itself from the hard drive and redirecting it to the output.
Try it some time. It's not trivial. And a self-replicating nanobot is,
essentially, a hardware quine.)

> But my key point is that insurance underwriters will not sign off on
> nanobots that are only proven the way drugs are generally proven today,
> via trials that show relationships.

Why would they do for for drugs in the first place, then?
And please, talk about replicating devices if you wish, but talk also
about non-replicating designs in you answer, if you would. I'm
specifically asking about non-replicating designs.

> Chemicals have been around for
> billions of years and we know they can be unsafe, but are unlikely to
> turn the planet in to a mass of grey sludge. This risk in
> nanotechnology requires very very very very detailed design fully and
> entirely understood, IMHO.

Chemicals have turned the planet into a mass of green sludge, actually.
And a mass of anaerobic sludge before that.

Still, it is NOT NECESSARY to design nanobots for self-replication.

--
John S. Novak, III
The Humblest Man On The Net

.

Relevant Pages

  • Re: OT: lets ignore bvallelys political postings
    ... ntelligent Design: The Origin of Biological Information and the Higher ... In a recent volume of the Vienna Series in a Theoretical Biology ... form first arose during the evolution of life. ... and classes of animals also arose at this ...
    (rec.arts.theatre.musicals)
  • Re: OT: lets ignore bvallelys political postings
    ... ntelligent Design: The Origin of Biological Information and the Higher ... In a recent volume of the Vienna Series in a Theoretical Biology ... form first arose during the evolution of life. ... and classes of animals also arose at this ...
    (rec.arts.theatre.musicals)
  • Re: Steven & Mark: Faith has an object
    ... Acceptance of material causation instead of Divine casuation refutes ... claims of evolution or the Bible. ... As explained over, and over, and over, the appearance of design is not evidence of either a designer, or "divine causation". ...
    (talk.origins)
  • Re: Mischaracterization of ID?
    ... > mischaracterization of what constitutes Intelligent Design. ... > creationism as it is an attempt to, using scientific methodology alone, ... evolution, nor those offered against it, but are convinced that ID is based ... Johnson), or progressive separate creations, or recent ...
    (talk.origins)
  • Re: How New a "Science" is Intelligent Design?
    ... Micro-evolution: evolution for which the evidence is so ... be able to prove design through exclusion of all alternatives. ... And the plants do their best to evolve ... The ID movement as a whole is an obvious creationist scam. ...
    (talk.origins)