On Replication

 

There is a growing interest in self-replicating machines. Beginning with the RepRap project, and now continuing into RepLab, the open source FabLab, there is a serious effort to build machines which can build themselves.

It is a laudable goal. A machine which can make itself can also make an unimaginable variety of other machines, each unique if desired, and promises an era of material abundance and freedom from scarcity. Pursuing that goal, however, has shown the goal itself to be somewhat unclear. This is an attempt to remedy this situation.

What is replication? The production of a replica, but what is a replica? The original meaning was that of a duplicate of a work of art, properly one produced by the original artist.

I propose we define replication precisely, as the transcription of a template into a physical form. This definition is perhaps more general than we are used to, but I believe it accurately captures the use of the word.

Let’s take the test case: an artist producing a replica of her own painting. The original painting may have taken many months, while the replica may take only a day or two at most. The artist is using tools, a brush and palette, and materials, canvas and paint, to make a copy of a template, the original work.

How about DNA self-replication? We have a toolchain of proteins, from DNA Transcriptase on, which transcribe a template, the DNA, into a copy of itself, using materials (ATP, other nucleic acids, caffeine if there’s some in the organism in question) both extrinsic to the cell and fabricated within it.

Note that in the first case the template resembles the result, while in second case, if we consider the result a new cell rather than new DNA, the result doesn’t resemble the template at all. This is immaterial to the act of replication; a high-resolution inkjet printer and a scanner could replicate the artist’s original work, faster and more accurately than she could, and in this case the template would be stored as bits within a computer.

The rest of this discussion will focus on replication in the context of human-tool interaction.

This means replication of an object, from a template, through the combined efforts of at least one human with at least one tool. We call this a ‘tool’ rather than a ‘machine’ for greater generality.

The earliest form of replication was a human using a tool to make a copy of an object. Our paradigm case for this is a scribe copying a scroll. A human can’t do this without the materials (paper, ink) and tools (quill, sharpener) required for the act. With those tools, a human can copy a scroll in a certain amount of time.

A few observations here: the tool enables, the human works. A faster tool, let us say a pen rather than a chisel on stone, allows for faster work, but more copies will always take more human time. We will call this Type 1 replication. It is characterized by modest gains in efficiency for mass production and poor scaling characteristics.

A second form of replication came about when humans began using tools to automate aspects of production. By investing time in setup, many copies can be replicated much faster. The work is still done by humans, with tools; but by doing the work intelligently, many copies can be made with comparatively low effort. We call this Type 2 replication.

The paradigm case for Type 2 is the printing press. By investing considerable effort in building a press, casting type, and laying out a document, one is rewarded by the ability to print as many copies as one wants, with speed and reliability which increase over time. Type 2 replication has good scaling characteristics and rewards mass production. It is characteristic of the Industrial Age.

Type 2 production next evolved in two directions the first of which we’ll call Type 2a: significant setup time followed by fully automated production. Newspaper presses work in this fashion, although the most recent ones are Type 3. Type 2a has significantly better scaling characteristics than Type 2, and mostly displaces it when developed.

Type 2b
replication is where there is significant tool investment, but negligible setup time and non-automatic production. The matching paradigm for this is a Xerox machine, though like newspaper presses these have become increasingly Type 3 over time. An early Xerox would make a copy of anything that fit on the glass, or even a hundred copies, but a human had to stand around feeding it sheets and doing things like collation by hand. This has poor scaling qualities, being labor intensive, but it favors customization over mass production.

Type 3 replication has negligible setup times and fully automatic production. Our paradigm case here is a laser printer. Even a desktop laser printer can print replicas of stored data, each different, for many hours without needing attention. The factory scale printers can make entire books without human intervention. Type 3 replication is characteristic of the Information Age, or Industrial Revolution 2.0.

Type 3 has good scaling characteristics and rewards customization over mass production. One can make an arbitrarily large number of copies off any given template, but there’s little to no advantage over using many different templates.

These are the essential distinctions:

Replication Type
Setup Time
Production Time (Human)
Type 1
Minimal
Variable and Significant
Type 2
Significant
Constant
Type 2a
Significant
Minimal
Type 2b
Minimal
Constant
Type 3
Minimal
Minimal

Many FabLab technologies, such as laser cutters, hobby-scale 3d printers, and CNC mills, are Type 2b. What’s interesting about 2b replication is that it’s often a small step away from being Type 3, needing only, for example, a small conveyor belt. I hope that by drawing attention to the radical difference in scaling quality between Type 2b replication and Type 3 replication, I can encourage open-source hardware developers to strive for Type 3 tools whenever possible.

A Type 2b tool is a creative enabler, while a Type 3 tool is both a creative enabler and an economically disruptive force, because it can compete effectively with 2a mass-production technologies. It also frees human labor for other pursuits, which is an important goal for many.

So this is replication; what of self-replication? This is the second concept I hope to make clear with this essay. Self-replication is not different from replication, except that, in the context of human-tool interaction, that which is being replicated is a part of the tool itself. I refer to this as recursion, because that’s what it is.

Note that there are several Type 1 replicators with a high degree of possible recursion. The lathe is a classic example: one can hand-turn many of the components for a lathe on itself, and even bootstrap a lathe by building the spindle and using it to make other components. A kiln is another example, since a kiln can be used to fire refractory bricks to make another kiln. In fact I might argue that a kiln exhibits the highest recursion of any Type 1 replicator, in that a kiln can be nothing but stacked bricks and a fire, and the bricks can all come from a kiln.

Note, however, that this still has poor scaling qualities, though it favors custom production. The Dave Gingery approach of contributing massive human inputs to a bootstrapped machine shop is noble, but it will never be an economic force. It is simply too easy to identify the main shapes needed, turn them over to Type 2 or 2a production, and swamp the market.

The RepRap project faces exactly this dilemma. RepRap knocked the bottom out of 3-d printing, but it did this mostly by making the MakerBot possible. The RepRap is a 2b replicator, which requires human tending during the entirety of the recursion process. This makes it easier and cheaper to simply make a chassis on a faster 2b tool, namely an Epilog laser cutter that NYC Resistor happens to already own, and sell that. If the RepRap were a Type 3, with an automatic conveyor, it might be a different story, since a RepRap could then print its chassis parts at a rate of a few complete sets a week with minimal human inputs. For that matter, a MakerBot with a conveyor belt would be a Type 3 replicator, albiet one exhibiting less recursion than a Darwin or a Mendel.

Let me conclude by suggesting that the RepLab project needs to place priority on Type 3 replication, not on a maximum level of recursion for the whole system. Recursion, I think I’ve shown, is no great trick if it takes a human hand-holding the machine through the whole process, since this is how machine tools have always been built, on other machine tools.

The power of Industrial Revolution 2.0 lies in developing machines which exhibit Type 3 replication. If the outputs of those machines are sufficiently general than the toolset as a suite can exhibit economically significant recursion. An example would be an open-source toolchain for automated printing and assembly of PCB boards. Such a suite would be able to print its own control hardware in addition to any other circuit that can be designed and mounted with a pick-and-place, which is most of them.

The RepLab project should focus on tools exhibiting Type 3 replication for a variety of economically significant goods. Recursion is an emergent result which can lend exponential momentum to the deployment of RepLab technology. By approaching the project goals in this order, we have a rational approach to an open source hardware platform for Industrial Revolution 2.0.

Advertisements
Published in: on November 22, 2009 at 10:45 pm  Comments (3)  
Tags: , , , , , , , ,

The URI to TrackBack this entry is: https://mnemnion.wordpress.com/2009/11/22/on-replication/trackback/

RSS feed for comments on this post.

3 CommentsLeave a comment

  1. response here:

    http://groups.google.com/group/replab/browse_frm/thread/bba7bd0881b10fc1

    – Bryan

  2. […] speak instead of recursive manufacture, and general replicators. Recursive manufacture is simple: it is building a widget using the widget. It is important, but […]

  3. People are already working on automated conveyor belts. http://charlespax.com/2010/04/27/motorized-conveyor-belt-sneak-peek/ and they’re working pretty good.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: