Architecture and Hardware

AI Xenobots

Xenobots carve traces through a field of particulate matter.
Researchers say xenobots are a completely new life-form, built from the cells of an African frog.

U.S. researchers using an 'evolutionary' artificial Intelligence (AI) algorithm say they have designed a completely new life-form, built from the cells from an African frog.

Dubbed xenobots, the microscopic living robots were designed using AI to complete a simple task: move in one direction.

While simple in function now, the xenobots can be enhanced to perform much more complex work with just a bit more tweaking. "We can imagine many useful applications of these living robots that other machines can't do, like searching out nasty compounds or radioactive contamination, gathering micro-plastic in the oceans, traveling in arteries to scrape out plaque," says Michael Levin, who directs the Center for Regenerative and Developmental Biology at Tufts University.

The research, which used AI to design a blueprint for the new life-form that was later assembled by a team of micro-biologists, is a first, according to Sam Kriegman, a Ph.D. student specializing in evolutionary robotics, synthetic biology and soft robots at the University of Vermont. "Ours are the world's first AI-designed organisms," Kriegman says.<

Leonardo Morsut, an assistant professor specializing in stem cell biology and regenerative medicine at the University of Southern California, agrees. "The truly novel aspect of this research is that it closes the loop between an evolutionary algorithm in silico and its implementation using living cells," Morsut says. "We have never seen this before."

According to Kriegman, one of the great advantages of using AI was its ability to generate and sort through "billions of candidate organism designs" in a search for those best suited for locomoting in a single direction. That enabled the research team to refer those preferred designs to microbiologists on the team, who used them as their guide to patch together the cell robots.

For software, Kriegman says researchers used a combination of VoxCAD and Python to develop optimum designs for cell robots capable of one-way locomotion, which the researchers wanted to build from heart muscle and skin cells drawn from embryos of the African frog Xenopus laevis.>

The open source AI code is freely available at The team has also made available background information on how the experiment was implemented at

Meanwhile, software for the experiment was run on the University of Vermont's high-performance computing (HPC) facility, according to Josh Bongard, a professor of computer science at the university.

According to Kriegman "Each independent evolutionary trail ran on a single CPU with 24 threads. The University of Vermont let us run 100 independent trials in parallel on 100 separate CPUs." In practice, "we simply supply building blocks and then tell the AI what we want the organism to do," Kriegman says. Each trial took 10 seconds to run, according to Kriegman.

Once the AI forged final designs for a robot capable of one-way locomotion, a microbiology team led by Levin used tiny forceps and a tiny electrode to patch together frog embryo skin cells and heart muscle cells to create the xenobots. The heart muscle cells, which auto-contract, gave the xenobot the muscle it needed to move in one direction.

Ultimately, the resulting cell robots developed were a success: they were able to use their own locomotion to move in one direction, according to Bongard.

Currently, the research team has no plans to commercialize their cell robots, but they do see great promise for the bots' future.

"We see this is a basic science result," Bongard says. "We hope that other groups will fork the code from our repository and adapt it for other machines that are partially or completely built from biological materials."

While a remarkable achievement in AI-designed robotic life, xenobots are still far from rivaling even the simplest life forms found in nature, according to Jean-Baptiste Mouret, directeur de recherché at Inria/LORIA who specializes in designing highly adaptive robots using machine learning and evolutionary computing. "We are still very far from the sophistication of frog, fish — or even a worm," Mouret says.

He explains animals have billions of cells organized in very specific ways, a design that is many orders of magnitude beyond xenobots.

Dylan Shah, a Ph.D. student at Yale University specializing in soft robotics, agrees. "Complex life forms often have a wide range of cell types that coordinate to create system-level organization;  for example, forming nervous and circulatory systems. Likely, we will need to figure out how to program algorithms which can generate and build such subsystems, and then integrate such subsystems."

>As for the ethics of using AI and bleeding edge microbiology to metaphorically 'duct-tape' together a multi-celled robot, Stephane Doncieux, a professor specializing in evolutionary computation, learning, and robotics at Sorbonne University says there's really not much to be concerned about. Granted, the research team's work creates life-like systems, he says, but those systems "are not alive, per se," Doncieux says.  "There is no nerve system, no reproduction system, no cell maintenance or food-supplying system."

Mouret agrees humans have been manipulating the building blocks of other life forms, like wood and yeast, for centuries, without any consideration over the attendant ethical implications. "Overall, assembling cells together to perform a specific task does not seem that different."

Joe Dysart is an Internet speaker and business consultant based in Manhattan, NY, USA.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More