“But then people argue, what's life?”
A new paper from the team behind the world's first living robots gets into the big questions.
Last month I published a piece with Pioneer Works about Xenobots—living, programmable micro-robots made from embryonic frog cells.
Developed by an interdisciplinary team at Tufts, Harvard, and the University of Vermont, Xenobots are the size of a quartered poppy seed. They live only ten days. They emerge from an eerie realm between cells and silicon—they’re designed by a genetic algorithm and refined in a computer simulation before being hand-carved from living frog goo and set loose in the real world.
And they do get loose. They swirl around the Petri dish, swarm, heal themselves, and gather particles—behaviors that are not well understood, even by their makers. As such, Xenobots are living models for understanding how cognition works on a cellular level. One implication, according to the team’s lead biologist, Michael Levin, is that cognition isn’t a binary, but a continuum, one that reaches down the evolutionary ladder and deep into life’s building blocks. As I wrote in the Pioneer Works piece, “the bad news is that there’s nothing especially magical about our brains. The good news is that there’s something magical in everything else that’s alive.”
Earlier this week, the team behind Xenobots published a second paper, in Science Robotics. This time, they streamlined the process, simplifying the Xenobots’ form factor by cutting out the genetic algorithm and allowing the Xenobots to cohere themselves into living spheres of cells. Rather than being actuated by the pulses of a frog’s cardiac cells, these more elementary bots swim around on the tips of the cilia that form naturally on their surface. They quite easily pass through tricky openings and mazes, and they seem compelled to collect debris and arrange it into piles, emergent behaviors that can be tapped—even, someday, programmed—to put biodegradable, self-powered robot swarms to work. In the future, a Xenobot might travel through your body, delivering targeted medicines, or float along natural waterways, conducting ecologically benign remediation.
On Friday, I spoke to three members of the Xenobots team—roboticist Sam Kriegman, from the University of Vermont, and biologists Doug Blackiston and Emma Katherine Lederer, from Tufts. When we spoke, Kriegman was coming in hot from a Twitter argument, in which a group of biologists were questioning the legitimacy of calling Xenobots “robots.” The robotics people, he says, have accepted the concept of biological robots; it’s a younger field, he suggests, more open to new ideas. “Robots just haven't worked for so long,” he adds, “and we’ve had so many failures.” Biologists are more resistant to the semantic overhaul. The trolls on Twitter called it hype.
To be fair, it is weird to call clusters of frog cells robots. It’s weird even for scientists, clearly. And in the absence of the usual cultural signifiers—metallic arms, cables, motors—what is a robot, really? Even within the Xenobots team, opinions differ. Kriegman tells me a Xenobot is a robot “because it's doing useful work, or something that is the rudiments of useful work.” Blackiston thinks it’s more a question of approach: he sits at the table in the wet lab, listening to classical music, assembling. “What I do is just like a roboticist,” he says. “I have little containers, except mine are cells instead of components, but I have sensors, I have actuators, I have scaffolds.”
So a robot is something that has a function, and it is something that is made. But don’t we have a function? Aren’t we made? It seems the more vital the concept, the more impossible it is to define. The team is batting around four different terms, internally, but none seem to get at the essential point. Blackiston proposes “living robots.” Kriegman sighs, “But then people argue, what's life?”
What’s life, exactly. Roboticists seem to think life is what it does. Biologists seem to think life is what it is. Perhaps the truth lies somewhere in between. “We've seen disciplines slowly collapsing to great effect,” Blackiston says. Still, biology is only newly acquainted with robotics and computer science. Their convergence is inevitable and will be fruitful; we are already seeing biological computing, evolution in simulation, and the beginnings of what Blackiston calls “the algorithms of life.” In the near future, he says, we may extract the mathematics from interesting biochemical interactions and use them to refine AI systems. Nature abounds with elegant solutions. Xenobots, then, emerge at the intersection (or collapse) of disciplines like AI, Artificial Life, cellular biology, and robotics.
They’re also the latest in an existing lineage of “biohybrid” robots. In 2016, a team at Harvard created a small stingray from rubber, gold, and the cardiac cells of a rat; its translucent rubber body pulsed like a heartbeat, propelling the chimeric creature through water (it was so beautiful it graced the cover of Science). That same year, a team at Case Western Reserve University produced sea turtle-shaped robots actuated by sea slug muscles. Since these hybrid designs require synthetic materials—rubber, polymer, a skeleton of delicate gold—they don’t benefit fully from nature’s innate plasticity, or its resistance to entropy. They’re unable to regenerate or self-repair. Cut a Xenobot nearly in half with a cauterizing knife, however, and it will put itself back together in 5 minutes flat. That’s some advanced engineering.
Living cells come pre-loaded with sensors, effectors, and signaling circuits. “If you build with living materials, you get a lot of things for free,” Kriegman says. Their onboard biochemical, biomechanical, and bioelectrical communication and control machinery is more effective than anything ever designed by a roboticist—3.7 billion years of evolution is quite a head start. I’m fascinated by this: on the one hand, we’re talking about something very new, something so new as to be practically undefinable. On the other, we’re talking about the oldest thing in the world. Although an enormous amount of craft goes into creating these organisms, it feels like a misnomer to say anyone built them. They built themselves, as we built ourselves, from the first split cell. What we’re talking about here is life itself, and what life can do.
More thought fuel for your weekend:
This interview with the science fiction writer Ted Chiang is worth every second of your time. Asked whether he believes we can (or should) design artificial intelligences that are moral agents, Chiang says, “long before we get to the point where a machine is a moral agent, we will have machines that are capable of suffering. Suffering precedes moral agency in the developmental ladder…in the process of developing machines that are conscious and moral agents, we will be inevitably creating billions of entities that are capable of suffering. And we will inevitably inflict suffering on them.”
A slower burn, but I loved this 1996 lecture by the architect Christopher Alexander—co-author of an iconic index of human-oriented architectural plans called A Pattern Language. Alexander was hugely influential in the world of Object-Oriented Programming, which is his audience here. He ends his lecture by encouraging programmers to think of software as the “natural genetic infrastructure” of an “alive, humane, ecologically profound” world with a “deep living structure.”