Self replication

From apm
Revision as of 17:43, 14 September 2017 by Apm (Talk | contribs) (General: some cleanup - needs much more)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
Highly compact self replication (starting almost from individual atoms)
is NOT a necessary requirement for the bootstrapping of advanced productive nanosystems.

Classification of degree of self replication

Self replication is not a yes/no question its more of a continuum of self replicativity.

Weak self replication

A possible definition of self replication in a very broad and weak sense:
The assembly of assemblers that assemble identically copied assemblers out of a set of base parts (sometimes called vitamins) that is equal or greater in number than two.

This definition is so weak that even a simple pair of pliers that can be used to put together another one out of two parts could be considered self replicating.

Since replicator mobility is not a requirement here exponential assembly falls under self replication in this broad sense.

Strong self replication

A possible definition of self replication in a very narrow and strong sense:
Self replication needs to fulfill all the five requirements of the replication pentagon

If the assembled product assemblers are not identical but rather randomly or autonomously mutated then using the term "reproduction" is more suitable and the replication pentagon extends to the reproduction hexagon

Strong self replication (actually self reproduction) can be dispersed over a wide system like e.g. human industry as a whole.

One can distinguish between:

  • compact self replication
  • dispersed self replication

Why compact self replication is not required for bootstrapping advanced productive nanosystems

Putting together a macroscopic object (consisting out of some 1023 atoms) almost atom by atom is a goal of AP Technology. It would take unfathomable amounts of time if it where done with only one robotic device. Massively parallel assembly is thus a necessity.

An early idea to solve this problem was an analog to biological cell growth. This idea naturally suggested itself so it was bound to emerge. This way self replication came into the focus as a possible pathway towards advanced atomically precise manufacturing early on.

It turned out that strong and compact self replication (a core concept of the direct path toward gem-gum-tec) is:

  • very hard to reach,
  • would lead to inefficient systems and
  • is undesirable.

Undesirable not because fear of bad image but because when the eligible but overblown fears (spawned from the fact that biology does strong and compact self replication seen in bacteria and viruses) are toned down to real levels there is still some worrisome material left.

Strong self replication with mutation in fact is an unconditional requirement for technological progress (tools making better tools) but it can be highly dispersed in subsystems.

So the problem can be split up into top-down methods bottom-up methods and possibly exponential assembly wedged in between. All of these individually are not compact self replicative. But all of them are part of the current human macroscale technology base which is self replicative (and self reproductive). (the incremental path toward gem-gum-tec)

See here for an overview over all available methods for bootstrapping massively parallel productive nanosystems: bootstrapping productive nanosystems

Exponential assembly

First few steps of partial structural replication via so called "exponential assembly". SVG

Exponential assembly is a method for copying/replicating structure that has the same exponential speed-up as self replication.

The units on their own lack functional completeness (they can not move individually) and the possible range of structural replication is thus limited to the working area.

It is even simpler than compact robotic self replication (in its usual sense) so the term was introduced to more recognizably distance the method from the usual naive associations that usually come with the mere mentioning of compact robotic self replication.

Exponential assembly may or may not be used to glue together top down and bottom up bootstrapping approaches.
Exponential assembly could be nested in a hierarchy. This may be a bit far fetched.

Alternate term suggestion: partially replicating assembly / partial self replication / immobile self replication

Current state of compact robotic self replicating systems

It was and still is widely believed that physical self replication requires systems of enormous complexity (state 2017). The goal of compact robotic self replication (which is not involved in the incremental path to gem-gum-technology!) was and still is often compared to the artificial recreation of life. This is a very bad analogy. It brings all five sides of the reproduction hexagon to the table when actually only five (remove adaptivity/mutation) or less (remove replicator mobility and building block availability) are needed for self replication.

It's still barely known that compact robotic self replication has already been demonstrated. (There's a link to video of the self replication process at the bottom.) Some "vitamins" like motors are used in the system but the system fulfills what one intuitively would expect from a compact self replicating robotic system. It's a concrete physical demonstration that compact robotic self replication is possible in a system with rather low complexity. Far below the complexity of say a current day operating system.

This work was done by Matt Moses in 2014.


Self replication can be very simple depending on which building blocks one takes for granted. The following start a simple self replication process with only the state of the building block changing:

  • A finger-tap to a chain or widening cone of standing dominoes. Replicating the fallen over state.
  • A crystallization core in super-cooled water spawning the crystallization.
  • A lifting off bird in a sitting swarm causing the whole swarm to lift off.
  • Fire? No fire is a bad example. What replicates there is not structure but destruction. Often complex molecules get converted to usually simple "ash" molecules. The products do not use their structure to form more products just their kinetic energy.

The next more complex form of self-replication is a composite unit that cause inactive building blocks to form more active composite units. [Todo: Link certain Video]

The human industry as a whole is a so called autogenous system. A set of many specialized assembled parts can collaboratively (and in complex sequence) create an equal set out of a set of base parts (ores). A complete AP small scale factory will be an autogenous system too.

In mold making one could in principal use two two-part-positives which where made from a two-part-negative to create a four-part-negative. This structural replication with parallel common guidance is called exponential assembly.

Classification based on base-structure size

  • Compact self replication with individual atoms as base material (the pure breed direct path) is obsolete.
  • Compact self replication with blocks of simple geometry self assembled from some foldamers as base material might or might not be involved in bootstrapping via the incremental path.
  • Compact self replication with crystolecules as base material will very likely be simple to implement due to its simplicity. Not for bootstrapping though rather as experimental product of advanced nanofactories. Note that the underlying stage where the crystolecules (which here are the "vitamins") are mechanosynthesized, is not so compact. At the crystolecule level (or even microcomponent level) assembly is done in a freely programmable general purpose way anyway. preproduced vitamin based self replication at the nano and microscale poses much less efficiency loss and a little less capability loss than the massive losses a molecular assembler working with individual atoms would have.

Block based self replication

A less top down alternative for exponential assembly would be block based self replication (using e.g. structural DNA nanotechnology). traits:

  • The robotic units consist out of simple basic blocks that can bind together. (complementary shape?)
  • The robotic units as a whole must be complex enough to fulfill their task.
  • A proto-robotic-unit (mechanism/linkage) must be assembled "manually" from the blocks.
  • Steering could be done e.g. by local broadcasting electro-statically from a chips surface.
  • There must be a method to feed the units with new blocks. (bulldozing & shape checking??)

Diamondoid self replication

Nanofactories of technology level III will as a whole be capable of doing diamondoid self replication via mechanosynthetic assembly of moieties as building blocks.

The original idea to make APM a reality was to build a diamondoid nanomachine of technology level III capable of self replication also known as molecular assembler. The attempt to directly build a proto-assembler with just a single AFM/STM microscope forces one to pack the whole replicative functionality into a very small package. This would make the unit inefficient. Furthermore the direct mechanosynthetisation of bigger structures necessary for a proto-assembler turned out to be a too steep slope without stepping stones (at least till the point of this writing 2014). There seem to be much more starting points for incremental technology improvement instead.

The idea of Assemblers blown up by the SciFi writers movies & co raised rather uninformed public concerns about runaway assemblers wreaking havoc. For the science community the "nano" tag meant/means(2014) funding money. But nano came with the meaning of APM embedded which they had nothing to do with. Then APM became linked with the (actually bogus) killer-nano-bugs. It seems some wanted to get rid of that (publicly as direct perceived) association. The prejudice of infeasibility from focused technical expertise may have played a role too. This culminated in the removal of anything APM related from the American national nanotechnology initiative NNI and drastic funding drop for APM development [TODO: check this].
See: history page for more details - Also see: Reproduction hexagon.

Classification based on the volume of a self replicating system

  • A (now obsolete) diamondoid assembler was supposed to fit into a box with a side-length of a few hundred nanometers.
  • For a self-replicative "partially assembled foldamer soup" (with thermally / non-positionally pre-produced parts as "vitamins") it seems hard to give a volume.
  • The size of a self-replicative "nanofactory pixel" could strongly vary with the degree of advancedness of the system.
    When crystolecules are treated as the vitamins it could be pretty compact (wild guess 32µm scale) but that might not be too sensible.
    When microcomponents are treated as vitamins (plausible) then it will likely be pretty big in relation to the nanoscale. (wild guess 1mm scale)

Macroscale self replication

Motivations for self-replication of macroscopic systems (no atomic precision involved) include:

  • Space exploration ("von Neumann probe" "astro-chicken") – space has the same inaccessibility problem of the nanoscale.
  • OSHW 3D printing philosophy (as propagated by Adrian Bowyer) -- providing enabling production capabilities to the masses
    Note that self replication goes only half of the way yet. 3D-printing can be seen as an (without plastic remelt) irreversible "pre-assembly" of vitamin printouts for the next step (note: normally only the non printable parts are called vitamins). This next step is a reversible assembly that is still done by human hands (state 2017).
  • Demonstration of semi-compact nanoscale self-replication (crystolecule level) – The "RepRec" project.

In contrast to nanosystems in macrosystems the needed material volumes quickly rise and this quickly become expensive. Thus at the macroscale there is a strong motivation for ultra-compact self replication. On the tight budget of "the masses" it is barely possible to afford even one printer and one assembly robot. In this configuration the printer cannot produce fast enough to feed the assembly robot. It is a severe bottleneck. Or the other way around: the assembly robot is severely underloaded.

(This very nicely illustrates a major problem with diamondoid molecular assemblers in the nanoscale where too the lower levels are much slower.)

In contrast to producing new stuff recycling of already pre-printed parts is super fast though. This is why one wants recyclability and part generality. Part generality not to the point of one single standard interface though. Since this makes the products unnecessarily bulky and isotropic (LEGO like - in a bad way). Too general parts lead to self replication only for the purpose of demonstrating the possibility of (macroscale) self replication. Too general parts lead to something that is effectively a "nonproductive replicator". Such a device is very impressive in its own right but this (to little public recognition) has already been for all practical purposes done. Check out the incredible demo by Matt Moses!

Much more desirable is a system with more specialized part types where some smaller part types shift the burden of standard interfaces to several specialized end effector adapters. Such a macroscale system (See: RepRec) could be truly useful for things like: 3D-printer frames, furniture, ...


For the attainment of technology level I either exponential assembly or block based self replication will be needed. Modular molecular composite nanosystems (MMCS) might be employed to organize self assembled structures of the upper size edge. The usage of standard blocks or other prebuilt AP structures for structural replication has the advantages that:

  • the needed accuracy is lower (click to place)
  • contrary to diamond mechanosynthesis no vacuum is needed
  • contrary to molecular moieties prebuilt structures can be stuck to a surface by drying and possibly cooling.

The other two methods for massively parallel assembly (or construction) known today are:

  • photo lithography for MEMS (not scalable to arbitrary small size scales; used in exponential assembly)
  • self-assembly (not scalable to arbitrary big size scales; used in block based self replication and possibly in exponential assembly)

For a more broad definition of self replication there is already a lot of literature to consult:
Wikipedia: Self-replicating_machine; The "Bunny Book": Kinematic_Self-Replicating_Machines; In general: Self-replication

(TODO: Make some notes about compact macro scale self replication based right from raw materials)


External links