What You Get When You Cross a Frog - With an Algorithm

The creation of “living machines” is a new, odd frontier in robotics

- Joshua Brustein

Michael Levin, a biologist at Tufts University, spends his days doing things such as coaxing flatworms to grow two heads or helping frogs regenerate lost legs. It may not seem like it, he says, but this counts as robotics research. In Levin’s view every organism is a type of machine, and he speaks with the vocabulary of a programmer, describing animals’ genomes as determining their “hardware” and explaining how he modifies living things by rewriting the “software” of their cellular communication and hacking their “bioelectric code.”

Some of Levin’s work involves biobots, small cellular blobs he and his colleagues build in a lab from the skin, muscle, and stem cells of frog embryos. His main collaborator is Josh Bongard, a computer scientist at the University of Vermont whose previous work has included building robots that could relearn to walk after, say, losing part of a leg. Although animals can recover from injury and adapt to new circumstances, robots and artificial intelligence systems still struggle to do so. The scientists are anticipating that combining elements of robotics and biology can change that.

The Defense Advanced Research Projects Agency (Darpa), the Pentagon’s moonshot arm, began funding Levin and Bongard’s work in 2018 as part of its Lifelong Learning Machines program, which seeks to apply biological mechanisms to novel computing systems. Levin uses a simulated version of natural selection to design his biobots—he calls them xenobots, for