[extropy-chat] Why no assembler design?

[Dan Clemmensen]

Hal Finney wrote:

>4. Designing an assembler is relatively straightforward and it is clear
>that it could be done today with a modest effort, but since there is no
>way to build the resulting device, no one wants to go to the effort of
>coming up with a complete assembler design.
>
>
>Answers 3 and 4 assume that merely designing an assembler is pointless;
>rather, effort should be devoted to designing an assembler which can be
>built from simpler tools, which is much more difficult.  Nevertheless it
>would seem that if the situation were close to case 4, it might be a
>worthwhile exercise just to make the technological potential more obvious.
>
>  
>
I've always assumed point 4: simple to design, but we do not know how to 
build it.

This changes the problem: dont merely design an assembler. Instead, 
design an assembler that can be built with simpler tools.

This in turn means that your problem statement is not quite correct. We 
should nto try for a complete design as a proof of concept and an 
inducement. Rather, we should be specifically trying to generate a range 
of assembler designs that may be achievable using simpler tools. If we 
keep generating valid assembler designs, perhaps we will find one that 
is buildable. Thus, we do not want the best assembler. Instead we want 
an assembler that can be built, no matter how poor this assembler is, as 
long as this assembler can build a better assembler.

Put it another way: the bootstrap problem is the only hard problem.

In particular, we do not need an sssembler that can operate in a hostile 
envoronment. We are free to pick any arbitrary environment for the 
bootstrap, as long as we can create that environment using macro- or 
micro- technology. Create the environmant, build the assembler using 
micro techniques, and then use the assembler to build a more robust 
assembler that can operate in more general environment.

For example, assume that the bootstrap assembler is built from nanotubes 
and that it will only operate in a hard vacuum, on a perfect diamond 
surface. Fine. We know how to create a hard vacuum and a diamond 
surface. If we can use this incredibly expensive system to build a more 
robust system, we win. I see a sequence of progressively more robust and 
sophisticated systems, culminating in a system that can build the 
critical elements of its own environment.

Here is a possible bootstrap sequence:

Nanotube assembler in a hard vacuum on a diamond substrate, with a 
highly ordered feedstock (benzene precipitated on diamond?) controlled 
and powered via an external laser and an external computer. This system 
may use MEMS structures to create nanotube feedstock.

Nanotube-based factory. Similar to the above, but with a number of 
specialized nano-machines to generate functionalized nanotubes as 
feedstock to the assemblers.

Diamondoid-producing nanotube factory. The nanotube factory can now 
generate a small set of diamondoid parts.

Nanotube factory with diamondouid tools. This factory can produce a more 
complex set of "crude" diamondoid parts.

Crude diamondiod factory: The machines in this factory are built from 
crude diamondoid parts, but can produce more sophisticated diamondoid parts.

Diamondoid vacuum factory: built from the output of the prior factory, 
This factory can build arbitrary diamondoid machines. Note that this 
factory and all preceding factories still depend on a hard vacuum and a 
highly-structured feedstock system delivered using macro-scale technology

 From here we can build a system that operates in a simpler environment, 
such as an argon atmosphere, and that needs a much cruder feedstock, 
such as methane or methane and carbon dioxide. We still depend on an 
external control system and an external laser power syhstem. We are 
still at the nano scale.

Micro-scale factory: This factory can synthesyze micro-scale components 
by connecting nano-scale parts.

Macro-scale factory: This is the first system that is (nearly) 
self-replicating. It can build  macro-scale containment, control, laser, 
power, and feed systems. It can also build just about any hardware of up 
to about 10cm^3.