Instant evolution

Evolution is the only known force capable of creating intelligent life, but it is not itself intelligent. Before animals could run, swim, climb and fly around our world there were billions of years of trial and error. This is because evolution lacks foresight, it cannot "see" ahead of time whether a mutation will be beneficial or catastrophic. At the Xenobot Lab, we developed an alternative algorithm that compresses eons of evolution into an instant, generating whole new species of self-moving machines—robots—from scratch by tracing failures in their behavior back to errors or inefficiencies in particular parts of their physical structure. A computational ("differentiable") fast lane that bypasses the traffic jams of evolution, the resulting machines are not shackled by the idiosyncratic events and frozen accidents that have shaped natural forms on Earth, nor are they bound by our imagination.

In press. 


Computer-designed organisms

Most technologies are made from steel, concrete, chemicals and plastics, which degrade over time and can produce harmful ecological and health side effects. It would thus be useful to build technologies using self-renewing and biocompatible materials, of which the ideal candidates are living systems themselves. At the Xenobot Lab, we design completely biological machines from the ground up: computers automatically design new machines in simulation, and the best designs are then built by combining together different biological tissues. This suggests others may use this approach to design a variety of living machines to safely deliver drugs inside the human body, help with environmental remediation, or further broaden our understanding of the diverse forms and functions life may adopt.

More info:

Kinematically replicating organisms

Almost all organisms replicate by growing and then shedding offspring. Some molecules also replicate, but by moving rather than growing: they find and combine building blocks into self copies. We discovered that clusters of cells, if freed from a developing organism, can similarly find and combine loose cells into clusters that look and move like they do, and that this ability does not have to be specifically evolved or introduced by genetic manipulation. At the Xenobot Lab, we use AI to design clusters that replicate better, and perform useful work as they do so. This suggests future technologies may, with little outside guidance, become more useful as they spread, and that life harbors surprising behaviors just below the surface, waiting to be uncovered.

More info:

Evolved virtual creatures

Current theories of evolution and cognition derive from a single data point: the animals and plants that surround us and their fossil record below our feet. Theories built on top of this data are essential in our quest to understand the natural world, but they may also lead to intelligent artificial systems with useful behaviors.  At the Xenobot Lab, we are continually expanding this vital dataset using computer simulations of other chemistries, creatures, ecosystems, and planets. Hundreds of millions of years of evolution can occur inside of the computer in a matter of days, yielding whole 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 it under different settings illuminates the environments in which particular cognitive structures and functions repeatedly arise and provides a glimpse of intelligent life as it might exist elsewhere in the universe.

More info:

Fractal robots

Ever since Euclid invented geometry, humans have engineered systems with simple shapes in mind: points, lines, planes, cubes, circles and spheres.  But these smooth shapes do not actually exist in the real world. Coastlines, rivers, trees, respiratory and vascular systems, brains and DNA all exhibit a nested organization of intricate yet self-similar structures, not unlike a matryoshka doll. If you break a branch off a tree and stick it into the ground it will resemble the original tree, only smaller. Such geometries, known as fractals, can be generated by simple rules yet hold unbounded complexity. At the Xenobot Lab, we explore the adaptive benefits of fractals, such as how self similar structure can, in some cases, result in self similar behavior. This raises the tantalizing possibility of mass producing a very small robot that carries out very small work (e.g. unclogging an artery) but may be combined in increasing numbers to form similar yet ever-larger shapes with similar yet ever-larger functions (unclogging a septic tank, dredging a river, terraforming a planet).

More info: