Nanotechnology Timeline

The Wayback Machine - https://web.archive.org/web/20060307220916/http://www.lucifer.com:80/~sean/N-FX/macro.html

Macroscale Replicator

Here's a proposal for a limited self-reproducing system that is far easier to build, with today's technology, than when Drexler is proposing, and almost infinitely easier than what the Los Alamos physicists are proposing, but is nonetheless challenging enough that I will wager $100 to the first comer (as well as funny money in the idea futures market, if somebody wants to set one up for this) that this challenge will not be met within the next ten years. (That's only 8 years before the supposed date of the Singularity, for those who are counting).

The milieu is familiar to many university researchers: Lego robots. I reserve the right to change the rules of this challenge until one month after a research project describe to me starts working to meet this challenge is actually funded; although I will be willing to fix the rules sooner within the scope of an idea future. I welcome suggestions to improve the rules.

The "raw materials" can be chosen from the standard separated parts in Lego(tm) elctro-mechanical kits sold in volume retail, or any other toy assembly kit (e.g. Tinkertoys) that has been sold at Toys R Us for at least one year straight, plus up to 100 additional _simpler_ parts subject to my approval. (I have played with Lego enough to know the tolerances may not be good enough for this sort of thing, so I'm willing for engineers to make their own better-fitting parts for the purposes of this challenge, but I don't want cheating by introducing complex parts specially designed for this task).

These toy parts may be disassembled into "rawer" parts by hand but may not be assembled into more complex parts by hand. Thus raw plastic, separated metals, computer chips sold with toys, sand :-), etc. may be used if desired.

Special firmware and software can be written for any computer chips so obtained, but no special electromechanical apparatus may be added as "raw material". Use of special-purpose chips not sold as a standard part of a toy kit is subject to my approval, and customized pre-assembled interfaces between chips and toy-derived electromechanical parts is subject to my approval.

The parts must be separated and set over a large area (I recommend a large gymnasium or an aircraft hangar). They can be sorted according to type (e.g., long red bricks go in this pile) but may not be sorted by assembly order or laid out in a particular orientation. There may be no other containers, machines, or other materials on the floor or interacting with the floor Lego machines.

The challenge: starting with less than 100 kg of any number and variety of machines pre-assembled from these raw materials, and the piles of separated parts, these seed machines will assemble over 10,000 kg of nearly identical duplicates of the seed machines, within 60 days. (I don't care if they "mutate" in < 1% of assembly configurations or parts content by the end, they just need to be substantially duplicates of the seed machines). No human interaction with the assembly process is allowed during this period, except in case of emergency in which case the process must be terminated and restarted from the 100 kg seed machines.

> We need to do [self rep] in normal scale before we can
> even attempt this at the nano level. (i.e. ordering a bunch of
> construction machines to build a bridge and maintain themselves to
> overcome any difficulties without human help)

In article [email protected], [email protected] (Tihamer Toth-Fejel ) writes:
>I did my thesis on self-test in self-rep systems. I was getting
>frustrated because my electrical engineering was turning into chemistry
>as my subsystems needed to check sub-subsystems, etc. When I talked to
>Eric Drexler about it (at) a conference, he said, "God has very good quality
>control on atoms."

The only existing one I know of is our entire Earth wide industrial complex. Nothing smaller than Earth scale, nothing bigger than molecular level.

I know there have been studies on this (NASA 1981, and more recently Discover Magazine October 1995 had an article about some guys from Los Alamos National Labs wanting to build a clanking replicator). Wouldn't it be useful to nanotechnology if someone attempted to build one? Not even a completely closed system. Draw the line at a certain component level. Any component below this line could be externally fed into the system. It seems there would be much to learn and much software to write in the process of doing this.

Even if the supplied components were higher level parts (e.g., wheels, block motors, battery cells, computer modules, etc...) I suspect this would be a very difficult project to make work.

If we can't do this, do we really expect the system control software to be easier to write for nano scale machines?

*-------------------------------------------------------------------*
| Anthony S. Napier LEXIS-NEXIS, A division of Reed Elsevier Inc. |
| Dayton, Ohio EMAIL: [email protected] FAX: (513)865-6808 |
| VOICE:(513)865-1864 WWW: http://www-dev.lexis-nexis.com/~anthony |
| "The best way to predict the future is to invent it." -Alan Kay |
*-------------------------------------------------------------------*

This could be an interesting project if scaled back to something similar to what we are looking for -- a self-reproducing assembler make out of a small set of building blocks using a limited set of operations: build an assembler out of erector sets/legos/ball and stick molecular models, tie it to a computer, and see if you can get it to reproduce itself.

I'd assume the computational side instead of building it (implementing rod logic in legos?), a well-defined construction environment (all pieces of one type in _this_ bin), and cheat with pre-fab motors and whatever other specialized devices are needed.

2. a kit that could be marketed to high schools for annual competitions to increase interest/awareness

On 28 Dec 1995, Jack Maus wrote:
> I'd assume the computational side instead of building it (implementing
> rod logic in legos?), a well-defined construction environment (all
> pieces of one type in _this_ bin), and cheat with pre-fab motors and
> whatever other specialized devices are needed.

This would indeed be a very educational project! It's a very ambitious one as well, I think. Nobody has ever built a Von Neumann replicator machine (which is basically what you're describing); it appears to be a horrendously difficult task. But perhaps, with the short-cuts you mention above, it could be done.

1. What will the building blocks be, and how are they joined? Something as close to atoms (in behaviour) as possible would be best, if we want our experience to generalize to molecular replicators. But basically, we should take whatever we can get to work: consider the amount of precision and force it takes to snap Legos together, or the complex operations needed to assemble an Erector set. Can we find or create something easier (yet still strong enough to do the job)?

2. Are the building blocks even capable of forming a "gripper" fine enough to manipulate other blocks? (With Erector, for example, I think the answer is clearly "no".) How should blocks be held -- by friction, magnetism, Velcro, what?

4. What is the assembler's environment: on land, underwater, or what? An assembler in a large fish tank might have advantages: we could mimic solution chemistry by having the building blocks neutrally buoyant, and stirred around randomly by currents. And since the structures would be essentially weightless, they might not have to be as strong.

I think questions 1 and 2 above are probably the most important for getting things started. Let's see how far we can take this.
,------------------------------------------------------------------.
| Joseph J. Strout Department of Neuroscience, UCSD |
| [email protected] http://www-acs.ucsd.edu/~jstrout/ |
`------------------------------------------------------------------'

Mar	MAR	Sep
	07
2005	2006	2007

	Milestone	Dependencies
Predecessors	N/A	N/A
Successors	Nanoscale Replicator	Macroscale Replicator to Assembler