Article: How Bio-Bots Are Making New Life

Photo Source: Flickr

The recent awakening of public awareness of artificial intelligence through the emergence of large language models such as OpenAI’s ChatGPT has at once popularized AI and simultaneously obscured another dramatic AI advance—xenobots. Over the past year, Sam Kriegman at Tufts University, researchers at Harvard’s Wyss Institute for Biologically Inspired Engineering, and the University of Vermont have used artificial intelligence to design biological robots that can produce generations of offspring. In their paper, “Kinematic self-replication in reconfigurable organisms,” published in the Proceedings of the National Academy of Sciences, these authors describe how they used a type of AI called a genetic algorithm to design different configurations of frog stem cells that can collect other cells to make babies in the likeness of the parents. (pnas.org)

The bio-bots described in Kriegman’s research get the name “xenobots” because they use stem cells derived from the eggs of the African clawed frog, Xenopus laevis. Earlier research demonstrated that the stem cells that lead to making frog skin can take different shapes with human manipulation. Additionally, frog skin has little cilia or tiny whips that move back and forth to distribute the slimy mucous that protects the frog from disease, bacteria, and fungus. In a laboratory setting, grouping skin stem cells in little blobs of the cell mass will make the skin cilia, which can flick back and forth allowing the blob of cells to swim around in water. Changing the shape of the xenobots optimizes them to perform different tasks. One shape described in the paper resembles a pizza with one slice removed. This shape allowed the blob of skin cells to scoot around and collect loose individual stem cells into new blobs that could mature into a new generation of pizza-like blobs and so on.

An essential part of the research into building robots out of living cells depends on an AI that uses the principles of evolution to design the best configuration of cells to achieve a desired outcome, such as reproduction. Such an AI called a genetic algorithm can start with a set of random shapes and run them through a simulation of the different shapes swimming in a petri dish with stem cells scattered around. The shapes that produced larger piles of cells got strong fitness scores, and those that did not were eliminated from the pool of shapes. Before the following simulation, some shapes get mutated to change their shape slightly, and the simulation runs again to find the best or fittest shapes. Repeating this process over and over finds the optimal shape. Genetic algorithms allow for a more efficient way to explore design without having to test every shape possible.

The type of replication of the xenobots represents a new mode of reproduction unique in the animal world. All organisms reproduce, and different species have different modes, such as sexual reproduction with a mother and father, as in humans, or parthenogenesis, which is asexual reproduction where a female can make a baby from an egg without fertilization. The xenobots rely on the presence of stem cells to reproduce, avoiding sexual or asexual reproduction. The significance of this finding emerges from its stark departure from evolution and having a machine use biological cells to design a “machine” to perform a specific task. The paper’s authors consider the potential of such tiny bio-bots for many tasks, such as organ regeneration, pollution remediation, and more. Additionally, the University of Vermont author Josh Bongard envisions a future where AI gets an actual body through which it can experience the world. (newscientist.com) Large language models like ChatGPT do not experience the world; they just use words to predict the next word in a sentence, but ChatGPT cannot know what it feels like to get wet. Bongard sees the future where AI can use living cells to experience the world. Such a future needs vigorous ethical examination before letting people, through AI-driven machines, reconfigure living things.

Dr. Smith’s career in scientific and information research spans the areas of bioinformatics, artificial intelligence, toxicology, and chemistry. He has published a number of peer-reviewed scientific papers. He has worked over the past seventeen years developing advanced analytics, machine learning, and knowledge management tools to enable research and support high-level decision making. Tim completed his Ph.D. in Toxicology at Cornell University and a Bachelor of Science in chemistry from the University of Washington.

You can buy his book on Amazon in paperback and in kindle format here.