Xenobots and anthrobots: what they are and are not

"Living robot" is an unhelpful phrase, and it is attached to some genuinely interesting biology. It is worth taking the two apart.

What was actually made

In 2020, researchers at Tufts University and the University of Vermont reported millimetre-scale clusters assembled from frog embryo cells. The clusters moved, and the way they moved depended on how the cells were arranged. Computer search was used to explore which arrangements would produce which behaviour before anything was built. The name that stuck was xenobots, after Xenopus laevis, the frog the cells came from.

A follow-up in 2021 showed that these clusters, moving through a dish of loose cells, could push those cells into piles that then became clusters themselves. Calling it self-replication stretches the everyday meaning of the word — nothing copies a blueprint — but the clusters did give rise to more clusters through their motion, which was not expected.

In 2023, a related result used human cells from airway tissue rather than frog embryos. These clusters, called anthrobots, also moved under their own power, and in laboratory conditions appeared to encourage repair across a scratch in a layer of neurons.

The part that is genuinely surprising

None of these cells were engineered. No gene was added or removed. The cells were taken out of their usual context and allowed to organise themselves, and they did something no one had asked them to do.

That is the real finding, and it is a finding about biology rather than about robotics. Cells carry more behavioural repertoire than their role in a body reveals. Remove the surrounding tissue and some of that repertoire shows up. What else is latent in ordinary cells is now a reasonable question to ask.

What they are not

• They are not machines. There is no control system, no programming in the ordinary sense, and no way to issue a command. Behaviour comes from the arrangement of cells and their own biology.
• They are not nanoscale. These are visible objects — a fraction of a millimetre or larger. Discussion of them usually arrives illustrated with pictures of nanobots in blood vessels, which is a different subject entirely.
• They are not medical treatments. The neuron result was a laboratory observation in a dish. Nothing here has been tested as a therapy in a person.
• They do not survive independently. They run on yolk or culture medium and stop after days or weeks.

The neighbouring field, briefly

Separately from cell clusters, there is long-running work on DNA origami: folding DNA strands into defined shapes, which has been done since the mid-2000s. By the early 2010s these shapes included containers that opened in response to a molecular signal, and a much-discussed 2018 study described DNA structures carrying a clotting agent to tumours in mice.

There is also work on biohybrid microrobots — a motile cell or bacterium harnessed to a synthetic carrier, often steered magnetically. Some of this is impressive. All of it is early. Despite two decades of coverage suggesting otherwise, no nanorobot or microrobot of this kind is in routine clinical use, and the hard problems are unglamorous: getting where you intended, being cleared safely afterwards, and manufacturing at scale.

Why we find it interesting

At Gataca we spend most of our time on data rather than on cells, so the appeal here is the epistemics rather than the robots. A search process proposed configurations, the configurations were built, and most of what was learned came from the gap between what was predicted and what the cells actually did.

That is the same shape as every useful computational result in biology: the model narrows the field, and the organism has the final say.