[extropy-chat] Re: Nano-assembler feasibility - technology

[Brett Paatsch]

> This post will be about technology.

Sorry for the delay, in getting back to this Chris, by all means
takes as long as you need to reply yourself (if you want to).

BTW. If you did not see Hal Finney's post to the ExI list
I think you should check it out.

> Brett Paatsch wrote:
>
> > On Sunday, March 28   Chris Phoenix wrote:
> >>...  Have you read Merkle's papers?
> >
> > I have read some. But none contained a spec with a full set of
> > parts for an assembler.   Is there one that does?
>
> No.

Okay. Thanks.

> But that's not necessary for making preliminary estimates of
> performance.

I think we are still talking past each other.

"Performance" of what? You don't seem to be referring to the
performance of a self-replicating assembler. So performance
of what?

> >>The computer may or may not be nanoscale, depending
> >>entirely on what's easiest.
> >
> > Sure attempts to design [..] a full set of components for a self
> > replicator at any scale would permit the chosen scale of the
> > design itself to be one of the degrees of freedom.
>
> What I mean is that the computer may or may not be at the same
> scale as the mechanical parts.

Okay, but I am not sure where you are going with this.

Wouldn't you agree that it is necessary, if a first assembler is
ever to be build in practice, that there would be prior to
its existence a specification for building it that includes a full list
of its necessary component parts and preferably some
way of putting them together?

>  The first few nanofabricators can
> be controlled by a large computer.  Then you can use a lot of
> nanofabricators to build the nanocomputer quickly.  A tabletop
> integrated almost certainly needs nanocomputers, or at least
> nano-logic. But the first, bootstrapped nanofabricators don't
> need a nanocomputer.

None of this seems unreasonable to me in so far as it goes.

> >>We have a specification that includes most functional
> >> components, and supplemental information about range
> > > of motion, required stiffness, speed, etc.  We know the
>> > functions it must perform.
> >
> > Having less than 100% of the necessary parts in the system design
> > for the first prototype amounts to having 0% of a designed (and
> > buildable) prototype.
>
> Why are you talking about prototypes?  Lots of pre-prototype
> engineering work can be done, and some of it has to be done
> before a prototype can be designed.

I used prototypes because I am at a loss for a better word. Prototype
captures most of what I wanted to say in that a prototype need not
be an optimised design and yet it can be sufficient to show proof
in principle. A prototype can be used to show feasibility and to
win buy-in from folks that might not have otherwise accepted.

Drexler in Nanosystems (I'm talking from memory) recognized that
doing design work in this space was not like doing design or
engineering work in other spaces. He coined the phrase "theoretical
applied science" to try and outline a discipline/approach for
working in that space.

Lots of time and effort could be expended producing specs for
parts that may or may not ever get used in the first assembler or
for anything at all. Engineering is not usually that much of an
all of nothing exercise.

If one cannot conceptualise the entire system to the level of
knowing what each of the individual parts that make it up
are then how could one hope to build it to any sort of
timeframe or to place an estimate on its construction ahead
of time? I think the answer is that one can't.

>  We knew we could get
> to the moon before we finished the Saturn V design.

I accept that that was so. But I don't think that is a comparable
example. I don't know much about the Saturn V but I can see
that the sort of complex systems designing in it might have been
approaches in a series of incremental steps with each step
representing a discrete product in its own right.

I'm thinking people at different times in history saw logs float,
then built dug out canoes, then boats, then submarines and then
a space vessel.

There were useful relevant examples and existence proofs
available to guide the exploration of the design space for
engineers good enough to see them.

But when I come to consider a self-replicating assembler
I do not get the sense that I am being asked to consider
something for which there are any relevant existence
proofs.

Robert used a term he said Mike had used Nano-santa.
I associated the term genie machines that I think Drexler
first used in Engines of Creation. These terms denote a
particular type of self-replicator. The sort that might
product not just themselves but also products like
diamondoid.

I know that no biological assemblers can produce
diamondoid.

There are no existence proofs that these sorts of
self-replicators assemblers can be made. And there are
no proofs that they cannot. We simply do not know if
they can.

>From where I sit neither the case that these sort of
assemblers are definately possible nor the case that they
are impossible has ever been convincingly made.

Both sides want to put the burden of proof on those
holding the other view.

Because I can see how attractive it would be to have such a
Nano-santa, or even a complete blueprint for one, I don't think
it is sensible to start from the presumption that it is possible
to build one.

The existence proofs for lesser types of assemblers are
not satisfactory to be also existence proofs for the much more
complex self-replicating assemblers.

>  So why
> are you focusing on the 100% completed blueprints?
> We don't have them, we can't have them yet, we don't need
> them.  Let's move on and talk about the interesting issues.

Because I suspect people (like me) would like to assume that
100% blueprints (full sets of specs for all the parts in assembler
number 1) are possible in advance of their having any intellectual
reason for asserting that they are.

I think it will only be when (actually if) we have the full set of
specs that we will have the relevant existence proof that the
self-replicator of the sort we would like to have is possible.

This side of that discovery I don't think work aimed at producing
parts can even be assumed to be progress at all. There is possibly
a big time trap there as seductive as the the search for the perpetual
motion machine.

> > If you don't know how many parts are needed to produce a
> > full design spec for an assembler then how can you know that you
> > have "most" of them?
>
> I said we know most functional components.  Not individual parts.  If I
> were talking about an internal combustion engine, I'd say we know it
> needs valves, a piston, a crankshaft.  But it's too early to talk about
> valve springs or crankshaft bearings.

I can't see any point earlier than the point at which a full spec is
produced when we will even be able to measure progress towards
that goal because prior to then I don't see how we can know
that that goal is in fact achieveable.

This is not to say that there is no value in nanotechnology or that
non-assembler nanotechnology cannot advance I think it can and it
will. But without the assembler it will be incremental advances only.

> > The question is is it possible? And the answer is we don't know
> > until we see a systems specification that shows a full set of
> > components necessary to produce an assembler (at any scale
> > would be a very good start)
>
> As far as we can tell today, from all of the science and engineering
> work that's been done, it's more than possible.  It should have
> excellent performance numbers.  Feel free to address the engineering.

I disagree. I don't think you know or anyone else knows that to
construct the assembler or a produce a full spec for it (as described
above) is possible.  Can you show me that I am wrong?

I think that you know, that you don't know that it is impossible altogether,
That is not the same as knowing that it is possible in any one or more
particular ways.

There is a position of agnosticism in the middle, that I think is the
logical
place to be at present, given that

a) we don't know that all the possible designs people will suggest will
be impossible.

and

b) we don't know that there is any particular design that is possible.

> But "We don't know for sure until we've seen it done" is interesting
> politically but trivial scientifically.  So I won't try to answer it here.

Can you show that it is trivial?   I am all ears if you can.

> > Without a list of components showing its possible any time and
> > cost numbers may as well be numerology "
>
> No, the time and cost numbers *may be wrong.*  "Numerology" is
> name-calling.  Time and cost projections may always be wrong.  If
> there's an undiscovered error in the engineering, they may be very
> wrong.  This is a given.  So why are you saying it in an engineering post?

1 + 1 = 3 is wrong.   Saying we are getting closer to the day when we
will discover the unicorn is numerology in the sense that I meant it.
We can't predict this side of the event when we will discover the
unicorn, or if.

I didn't mean to insult with name calling. I did mean to be provacative.
I did not know that you in particular would be provoked.

Regards,
Brett Paatsch