Sam Kriegman
Sam Kriegman is a computer scientist working at the intersection of evolutionary robotics, artificial life, and computational design, where his research explores how machines and organisms can evolve new forms through simulation and biological experimentation. He is known for his work on xenobots—living robots created from frog cells.
Essay
Sam Kriegman, From Galatea to Xenobots
Xenobots, synthetic lifeforms that Sam Kriegman and his colleagues invented in 2020 through computational evolution and bioengineering, defy ontological divisions between “life” and “machine.” The team used tissues from clawed frogs, Xenopus laevis, to imbue material behaviors first programmed using computer simulation. Xenobots’ most fascinating property is that, unlike every other animal that has ever existed before on this planet, they evolved, and are evolving, inside of computers before being brought to life in the laboratory.
Current theories of evolution and cognition derive from the animals and plants that happen to surround us and their fossil record below our feet. Theories built on top of these data are essential in our purely intellectual quest to understand ourselves and our place in the cosmos, but they may also lead to practical technologies that help combat climate change and alleviate medical suffering. The future of our planet and species thus depends on our continual expansion of this vital dataset.
Lucky for us, this critical task aligns with enduring efforts to not only sense and imagine life, but to create it de novo. For our ancestors, it paid evolutionarily to ascribe agency to inanimate objects. We are programmed to see and hear an agent lurking behind innocuous shapes, shadows, and sounds. We are always creating virtual creatures in our minds. We see their eyes and faces in car headlights and electrical outlets, their bodies in ink blots and in cloud formations.
The fascination to bring inanimate matter to life is an enduring human dream. The Jewish Talmud describes Adam, the first man, as a golem, a partially molded human figure out of mud. In ancient Greek mythology, Prometheus fashioned man out of clay. The Roman poet Ovid wrote how king and scholar Pygmalion fell in love with his carving out of ivory, Galatea, who by Aphrodite’s grace sprung to life. Humans have mimicked these divine interventions through play, stories, and metaphors.
Computers now allow us to not just speculate about pasts and futures of life but indeed to see them play out on screen. We can propose alternative chemistries, creatures, ecosystems, and even whole new planets to see what evolves. On a single consumer-grade laptop, hundreds of millions of years of evolution lapse in a matter of days, yielding wholly new phylogenetic trees grown root-to-branch before our eyes. Parallelizing this process across multiple computers allows multiple histories of life to unfold simultaneously. Restarting life’s evolution under different biogeochemical conditions illuminates the environments in which particular morphological and cognitive structures and functions repeatedly emerge. Patterns that hold across a wide range of conditions indicate common, if not universal, forms of intelligent life that might exist beyond Earth.
These early results signal that biology is at the dawn of a new episteme, as existing concepts of cognition, synthetic life, “the natural”—and even evolution itself—are no longer sufficient to describe a growing catalogue of concepts of what life will become.
The first virtual creatures were evolved by Karl Sims in the early 1990s. In doing so he helped ameliorate a 400-year-old Cartesian wound that divided body from mind. In a computer program, Sims allowed both the bodies and brains of autonomous agents to co-evolve so that they ran, jumped, swam, moved toward light, and fought head-to-head for resources in a virtual world that followed (more or less) the laws of classical mechanics. The creatures were relatively simple, composed of just a handful of jointed, rigid components—but their behavior was surprisingly rich and lifelike.
Sims’ experiments started with an objective such as locomotion and a population of randomly assembled creatures. Although it was unlikely that any of these initial creatures would fully satisfy the objective, by replacing the worst-performing designs with slightly and randomly modified copies of the better ones, the population made incremental progress, generation by generation. It was the survival of the fittest; or, in the case of locomotion, the fastest.
In 2000, Hod Lipson and Jordan Pollack transferred creatures like Sims’ from simulation to the physical world with a technology just then emerging: 3D printing. Robot designs were rapidly and safely prototyped as virtual creatures, discarding the truly awful or dangerous designs before testing them in reality. These robots embodied the evolved truss-like forms of their digital counterparts using a plastic skeleton of bars, fleshed (ball-and-socket) joints, and linear muscles (pistons) which generated their evolved pushing, hinge-like, and ratcheting motions. Lipson and Pollack needed only to specify these building blocks and snap the motors into the body; the rest evolved on its own. “Robotic lifeforms”—embodied machines—were designed, optimized, and built, end-to-end, with almost no human intervention.
Twenty years later, virtual creatures took their third step forward: they were brought to life. I evolved diverse candidate lifeforms as virtual creatures to perform some desired function, and the best designs were then created by my colleague Doug Blackiston using a cell-based construction toolkit to realize living systems with the predicted behaviors. This yielded a continuous flow of synthetic living organisms—“xenobots”—that performed useful tasks yet bore little resemblance, above the cellular level, to any existing organisms. Our data and knowledge of life were thus expanded, not just in silico (in the computer) but in vivo (within the living milieu) as well.
*
The resulting xenobots present unique challenges to several longstanding ontological divisions. They are composed entirely of living tissues but their anatomies and behaviors are determined by rational design. They have become known in popular culture as xenobots because their cells came from one-day-old Xenopus frog eggs (blastulae). By rearranging these cells into new forms, xenobots have been programmed to walk, swim, behave collectively, and manipulate small objects. In a xenobot, there is no distinction between the brain and the body. There is no neural tissue inside a xenobot, yet they are able to do some interesting things and can even heal themselves after being cut almost in half. This is possible because they are built from cells, and each individual cell is itself a complex, adaptive machine. Even if put together into new configurations, cells sense their environment, communicate with one another, and take actions. This leads to a form of collective intelligence, even if it is not neural intelligence.
Certain rearrangements of cells cause xenobots to build copies of themselves kinematically (through movement), a form of biological perpetuation never previously observed in nature above the molecular level. This mode of replication did not evolve under eons of specific selection pressure toward form and function, but rather appeared spontaneously in a matter of days. A swarm of these self-motile units, presented with a field of dissociated cells, pushed them into piles as a byproduct of their motion, and those piles matured into self-motile units in their own right. These data reveal a novel example of biological plasticity: wild-type cells without genetic modification can be shifted to a completely different form of replication than the species’ default. Our knowledge of life was thus expanded in a surprising direction, beyond how we know that natural forms evolve on Earth.
The xenobots’ mode of self-replication challenged yet another ontological distinction: that of sexual versus asexual reproduction. In nature, life follows one of those two reproductive pathways. But a single xenobot, five xenobots, or a dozen parent xenobots can all perform kinematic replication. If parent xenobots remain in the same dish as their children, they can work together in teams to build grandchildren and great-grandchildren, and so on. This forces us to rethink an ontology that previously seemed to faithfully capture an ultimate truth—that reproduction is growth from a piece of either one parent or two—but is in fact a proximate detail of the particular evolutionary trajectory we happen to inhabit.
These experiments also demonstrated how virtual evolution can automatically design parent shapes that amplify replicative ability, and how such systems could, in future, perform exponentially useful work as a side product of replication. The useful work to which these biological constructs may eventually be put—and the way in which they mechanically assemble their offspring—suggests they may instead be viewed as autonomous robots.
*
Whether we call xenobots robots or organisms, rationally designing and optimizing these autonomous, self-motile and self-replicating biological systems is not only strange but extremely non-intuitive. And the design problem is only going to get more challenging if future xenobots utilize genetic modifications that give rise to complex sensorimotor feedback loops, action alternatives, memory, learning, and other increasingly cognitive behaviors. To make the design problem tractable for human minds, the solution space would need to be winnowed down to a vanishingly small subset of possible forms and functions. Breaking free from design constraints imposed by human cognitive limits greatly widens our search for useful technologies and new knowledge, but it requires, by definition, non-human assistance. Virtual evolution, AI, and other computational tools are poised to help.
Computers were used to rapidly generate a diversity of buildable xenobot forms that maximized a desired behavior, and maintained that behavior across a range of simulated conditions. This not only reduced biowaste by filtering out billions of bad designs before we attempted to build anything in the wet lab, but it also led us into parts of design space where human engineers typically do not wander. There, we not only found creative xenobot designs (fractals, strange asymmetries, and porous structures) but also new biological design principles.
The etymology of xenobots’ prefix, “xeno-” (ξενο-) from the Greek, is a “guest, stranger, foreigner.” For the ancient Greeks, guests were sacred. Xenobots and their increasingly competent descendants may be forever banished from the fields of robotics and biology as strange clusters of cells that are not really robots nor organisms. Or, they may be invited as strangers into the realm of life, welcomed alongside other more-than-mechanical, other-than-biological forms pushing us to redefine nature.