Recently, the sci-fi dream of self-replicating robots has been in the news, thanks to the University of Vermont, Tufts University, and the Wyss Institute for Biologically Inspired Engineering at Harvard. A recent experiment with frog cells was hailed by news outlets as disparate as CNN (“World’s first living robots can now reproduce, scientists say”) and Daily Wire (“American Universities Create First ‘Self-Replicating Living Robots’”). And it was also debunked by Ars Technica: (“Interesting research, but no, we don’t have living, reproducing robots”).

So what’s really happening?

Self-replication is a very tricky problem of information. To truly self-replicate, an organism must completely copy the information necessary for function. Seems simple enough but it introduces a conundrum. For the organism to copy its information, it must also copy the ability to copy its information. Solving this problem of self-reference is very difficult.

The easiest realm in which we can solve the problem of self-reproduction so far is computer code. John von Neumann (1903–1957), one of the inventors of the modern computer, designed a self-replicator known as the Von Neumann Universal Constructor. The Constructor is a genius piece of insight, anticipating the discovery of naturally self-reproducing code in DNA.

However, this is not the same scenario with the xenobots that are making the news. They are robots in only the loosest sense of the word. The term “robot” refers to a complex mechanical and computational entity that engineers have carefully crafted. These xenobots are actually stem cells from an African frog (Xenopus laevis):

The biologists, Douglas Blackiston et al., did not create the stem cells. The stem cells already showed self-reproducing behavior, as is the nature of stem cells. The biologists contributed only the fact that a computer cranked through many configurations of these stem cells to find one that was a bit better at self-reproduction — and not very much better.

Calling these cells “self-reproducing robots” is like saying that humans create catbots when a pet cat produces a litter of kittens.

To add glitz to the paper, the researchers claimed to use an “evolutionary algorithm” to do the cranking. But the algorithm is not really evolution as we know it. The algorithm picks the best members from a set of randomly generated configurations and then further randomly varies each individual member.

Thus. the evolutionary algorithm assumes that self-reproduction already exists and that there is no genetic crossover or survival of the fittest via competition for resources. The only way in which the algorithm is like evolution is that it includes an element of random variation and selection — but the selection is for a predefined goal. It does not arise naturally from the algorithm itself.

However, the biggest assumption in the algorithm is that all the machinery that does the hard work, the stem cell, already exists within a simulation that the algorithm uses to evaluate each configuration. So, even though the use of the term “evolution” makes it seem that the researchers generated self-reproducing cells through the use of evolution, their work is not at all like the conventional biological model of evolution, which is upposed to start with nothing and have no designer picking the likely winners.

All in all, the research was reported in a manner that implies that the scientists had discovered a plausible way self-reproduction can evolve. While their work is a very clever achievement, the paper does not provide a plausible path for self-reproduction, nor do the jury-rigged organisms on display actually self-replicate.