Evolution Icon Evolution
Life Sciences Icon Life Sciences

Self-Replication? Not Even Close

Photo: Molecubes, via Wikipedia.

Mathematician Granville Sewell recently wrote here about a YouTube video from Cornell University researchers who allegedly “created a machine that can build copies of itself.” The video shows a string of 4 modular cubes or “molecubes” that twist and turn and pick up other molecubes to form a second string of 4 molecubes. Impressed by this result, a commenter on Stephen Meyer’s Facebook page asserted that, oh yes, humans are indeed close to building a self-replicator.

Begging to disagree, Sewell notes (“Oh-So-Close to Self-Replication”) that the molecubes are conveniently provided by the researchers, rather than being built during the “self-replication” process. With tongue-in-cheek amusement, he adds:

So I stand corrected: although human engineers are still not able to design self-replicating machines, we are oh-so-close. All that is left is to add a factory to the Cornell device so it can produce for itself the molecubes that humans keep building and feeding them. Well, of course then we would have to add a factory that could produce this molecube-producing factory. And then…

I’m definitely on board with Sewell’s point. After all, in Evolution and Intelligent Design in a Nutshell, I contributed a chapter on self-replication (“A Factory that Builds Factories that Build Factories That…”). The Cornell molecubes didn’t build themselves. Instead, they were built by intelligent researchers using other tools and systems — by a separate “factory” so to speak — that was, in turn, built by other tools and systems, and so on. Yet beyond the observation of this uncomfortable regress, there are several additional instructive issues we need to examine if we are to really appreciate what self-replication entails.

The “Self” Isn’t Optional

If we’re serious about self-replication, then we have to take the self part of the equation seriously. After all, all kinds of things can be replicated. If I find a widget and am able to create a copy by hand, I could end up with multiple copies, but it certainly wouldn’t mean that the original widget self-replicated. I could add tools to the process, perhaps replicating the widget more quickly and with more precision, but that wouldn’t mean we could ignore my involvement in the process and ascribe the replication to the widget. I might even go on to design and build a sophisticated automated factory, with identical copies of the widget rolling off the assembly line in rapid succession, but that wouldn’t mean that the first widget had self-replicated into multiple copies.

True self-replication requires not simply that the widget be replicated (note the passive voice) by an intelligent agent or engineer (whether directly or by using tools and processes), but that the widget replicate itself. In the one case the replication capability and process resides entirely outside of the widget; in the other case the capability and process takes place within. Although challenging in its own right, the former is quite doable and is something we see happening in human technology on a regular basis. The latter is a much more onerous task, one that’s still far beyond the capability of any human-built system.

This isn’t to argue that self-replication is inherently impossible. But we had better understand the scope of the task if we are ever to achieve anything close to this remarkable engineering feat. Further, we need to appreciate what self-replication actually entails if we want to avoid being misled by the imaginative storytelling and deceptively optimistic headlines from abiogenesis researchers and university press departments.

Back to Cornell’s molecubes, Sewell correctly notes that the molecubes were conveniently provided by the researchers and that we’d need to “add a factory to the Cornell device,” so it could produce the molecubes. Yet even if some other “factory” existed to automatically supply an endless stream of molecubes, the actions of the molecubes would still not constitute self-replication. Instead, bona fide self-replication would require building the factory as well, along with all the systems and components and parts making up the factory.

Homing in on the Parts

One of the pervasive mental mistakes made by many abiogenesis proponents is failing to think through how the parts come on the scene, lazily ascribing their appearance to sheer happenstance. Yet the provenance of the parts is a critical piece of the whole puzzle.

This is doubly true in the context of abiogenesis, which is, after all, one of the primary areas of interest for people making dubious claims about self-replication. There were no cellular parts conveniently lying around in the barren wasteland of the early Earth. The first self-replicator had to not just assemble a copy of itself with conveniently supplied parts, but it had to build the parts as well. Further, all of that capability and all of those processes had to exist and be driven internally, not by some outside external agent.

A little bit of spin comes with the story. In addition to the dubious claim that they had created “a machine that can build copies of itself,” the Cornell researchers decided to call the cube-shaped blocks “molecubes.” Whether it was the researchers’ intent with this term or not, a reader might be tempted to imagine some kind of parallel between these cubes and molecules — you know, molecules that might just be lying around in the environment on the early Earth. Readers’ minds might even be drawn to the video image of the molecubes coming together to form a chain, the same way amino acids are thought to come together to form a self-replicating chain in the RNA-world hypothesis.

Unfortunately, the linguistic phraseology obfuscates, rather than represents, the reality. The one aspect of the molecubes that bears some analogy to an amino acid is that each molecube has electromagnets on a couple of the faces that allow the cubes to stick together. One might argue that this ability to bind to one another (sticking together electromagnetically) is analogous to the way amino acids or other molecules can bind together to form lengthy chains. Perhaps, but that is where the analogy ends.

A System of Systems

In fact, Cornell’s molecubes are not like simple molecules or even somewhat more complicated amino acid analogs. Each molecube represents a carefully engineered system in its own right. Consider even a basic high-level description:

The housing and magnets are made of different materials. Each molecube is made of two halves that can rotate around a transverse axis, similar to a child’s snake puzzle toy. How is this rotation accomplished? Each molecube contains a micro-servo motor — itself a carefully engineered system composed of multiple parts and multiple types of materials. Even more notable, each molecube contains “the complete computer program for replication,” meaning, we must presume, that it contains micro-electronic information storage, retrieval, and processing systems. Oh, and let’s not forget the energy requirements. Either power has to be conveniently provided by an external power system, or when this toy is purchased from the molecube factory, fortunately “batteries are included.”

This is just a high-level sketch, and all of these subsystems are themselves purposely engineered and carefully constructed parts of the larger molecubes. Yet precisely none of them are replicated by the molecubes.

How Far Did We Get?

When carefully brought together by capable Cornell researchers toward a particular end, what do all these different materials and sophisticated physical and electronic systems do? They enable the molecubes to stick together and form a chain of molecubes. That’s pretty much it. It is interesting work and a nice demonstration of micro-robotics, but it has precious little to do with self-replication.

Self-replication would require that the overall system actually, well, replicate itself. Including its parts and pieces and different materials, and the information storage and processing systems that go along with it.

Thankfully, not all watchers of the YouTube video were so easily swayed. One astute commenter channeled his inner Inigo Montoya from The Princess Bride, observing, “I don’t think you know what ‘self-replicate’ means.” Another observed, “They aren’t replicating. They are magnets moved by a computer.”

In fairness, the Cornell researchers have not, to my knowledge, claimed that their molecubes have great relevance to biology or the origin of life. Their hope is that future robots working in distant environments, such as Mars, might be able to self-repair. This is a laudable goal and they have achieved some interesting results in micro-robotics.

Yet despite the fact that self-repair falls far short of self-replication on the engineering spectrum, the siren song of self-replication lingers in the minds of listeners and, as in this case, gets trumpeted as an argument for some kind of naturalistic origins scenario.

Not Even Close

Some might note that the Cornell video is several years old and assume that surely we’ve made substantial progress toward building a self-replicating entity since then. Not really. When we carefully look at what is actually required for a self-replicating entity, we realize that the dream of self-replicating technology is but a wishful hope on the distant horizon. As I noted just last year in Nutshell, in the context of claims about so-called self-replicating 3D printers:

If one is the trusting sort, it’s tempting to look at projects like RepRap or BI V2.0 and think, “Wow! We are almost there. We’ve nearly created a self-replicating machine!” But a closer look is warranted. Neither RepRap nor BI V2.0 is self-replicating.

Not in theory or in practice.

Not even close.

Not even in the ballpark.

Nail in the Abiogenesis Coffin

There is a final important point that has devastating implications for the theory of abiogenesis.

It’s one thing to acknowledge that humans have yet to create a self-replicating entity, and another thing to say it didn’t happen by chance on the early Earth. Abiogenesis proponents assure us that on the early Earth (or perhaps in the far reaches of space) a few molecules accidentally got together — somewhere, somehow — and formed the first self-replicator: “a particularly remarkable molecule,” as Richard Dawkins gushingly called it, with “the extraordinary property of being able to create copies of itself.”1

Upon the emergence of this very first ancestor of all biological life on the lifeless planet, the power of Darwinian evolution would then kick in, transforming the primitive self-replicator into the first living cell and eventually, with the help of countless copying errors and eons of time, into the vast complexity and diversity of life we now see teeming across our fair planet.

This is nonsense. Contrary to this simplistic and naïve materialistic creation story, when we understand what is needed for true self-replication and the incredible engineering required, we come to realize that the abiogenesis paradigm is fundamentally flawed at the most basic level.

As I wrote in Nutshell

Self-replication, contrary to the materialist abiogenesis story, is not the beginning feature, a rudimentary trait that a single molecule could handle. Rather it is a culminating trait, one of the most dazzlingly high-tech traits in the biosphere. The accumulated evidence, taken together, strongly suggests that self-replication lies at the end of a very complicated, deeply integrated, highly sophisticated, thoughtfully planned, carefully controlled engineering process.

Self-replication simply cannot be the first trait of biology, the trait from which all others flow. Rather, it is the ultimate trait — the end of an organism’s creation, not its beginning.

If that’s right, and evidence suggests it’s good money that it is, then the abiogenesis paradigm is broken from the beginning. It is not just a question of working harder or using different molecules or getting more funding or even coming up with new ideas about how abiogenesis might work. The abiogenesis paradigm is upside down and backwards, flying in the face of what we know, both from an engineering perspective and from observing the marvel of sophistication that is the simplest self-replicating bacterium.

The theory of abiogenesis doesn’t just need a tweak here or a refinement there. With its placement of self-replication at the starting point of the history of life, it is deeply and fundamentally flawed at the most basic level and needs to be discarded.


  1. Richard Dawkins, The Selfish Gene, 30th anniversary ed. (New York: Oxford University Press, 2006), 15.