Robotics experts have long marveled at the swarm intelligence of ants and termites working together toward common goals. Now, new research demonstrates that computer-designed, biologically based organisms known as xenobots could wield that power to serve humans.
Previewing a merger of the natural and virtual worlds, the research, published Tuesday in Science Robotics, has major implications for advances in biological design, regenerative medicine, environmental decontamination and even the basic idea of human-driven research, said Douglas Blackiston, a senior scientist with the Allen Discovery Center at Tufts University and the lead author of the paper.
"At a certain level, I'm working for an AI right now," Blackiston said in an interview with The Academic Times. "I'm building what a computer gives me a blueprint for. That's really an inversion of how we've thought about biology, of scientists driving science. We're seeing the emergence of a new way that AI is moving into bio-design."
The xenobots, built from frog cells, are about the size of a grain of sand and can survive in a dish of tap water for more than a week on their internal yolks (and even longer if given sugar-rich food). They're resistant to contamination because they produce an external slime, like frogs, and can tolerate temperatures from 4°C to 26°C.
A previous paper on the first living robots proved the concept in principle with computer simulations. And the researchers have now demonstrated that swarms of the xenobots can complete tasks such as gathering tiny particles, such as microplastics in waterways, repair themselves, record memories and move through different types of environments.
Blackiston said the idea started as a joke he made to researchers at the University of Vermont who had been running virtual simulations of soft robots made of silicon bubbles that could expand and contract.
"I just joked in a meeting and said, 'I bet I could build your model out of living cells. We could build a living robot. You design it in the computer and I'll build it,'" he recalled. "That was really the focus of the first paper, I developed that technique and was able to build a robot out of living materials. Once we saw what was possible, the floodgates opened. What would you build? Why would you want to do that?"
Unlike biohybrid designs that have produced swarms of similarly tiny robots, which have rigid skeletal structures made of metal or plastic, the living robots have no synthetic components and are 100% biodegradable. Propelled by thousands of tiny hair-like structures called cilia, they can also move through water — and, theoretically, inside an animal. And as biological structures, they can heal themselves with built-in wound repair, unlike, say, a Mars rover if it loses a wheel.
"It's amazing," Blackiston said. "In the soft robots community, the best bots can heal from maybe a small puncture wound, like a rubber tire that can sort of heal from a nail. With ours, you can almost bisect them in half, literally just flay them open, and they zipper back up and go back to what they were doing. That's a huge advance in soft-bodied robots."
And there are exciting possibilities in terms of sensing and processing surrounding molecules, he said: "If you put these in the water, could it metabolize pollutants in a way that a traditional robot never could?"
Though the team couldn't be specific about pending research, they have already moved beyond using frog cells to build xenobots. They're also seeking to make them more tolerant of saltwater in hopes of helping address the microplastics crisis in the world's oceans.
"If you happen to lose a couple, they just decompose, so they're not going to cause any more plastic or artificial buildup," Blackiston said.
For those concerned with doomsday scenarios caused by runaway, self-replicating technologies (see: the gray goo theory), there's no potential for the xenobots to consume the entire planet à la the cartoon robot Bender in a 2011 episode of Futurama. Though the xenobots can metabolize matter from their environment, frog skin cells cannot reproduce themselves, which serves as a killswitch built into the system.
"I have a really hard time coming up with a dark scenario for this," said Michael Levin, the director of the Tufts Center for Regenerative and Developmental Biology and a co-author of the study. "These things are painstakingly made in the lab and only exist for a couple of weeks. They biodegrade. Compared with the stuff other people are making in terms of genetically modified organisms that are able to reproduce and go out into an ecosystem, this is extremely innocuous."
Moreover, the researchers aren't releasing the xenobots into the wild anytime soon — they're still limited to a lab setting.
The xenobots' behavior can be dictated by a series of "if, then" statements, much like a computer: If you sense this chemical, then release a decontaminant. After two days, swim to the surface. If you don't sense this chemical, then disintegrate.
"We also gave them, with a molecular intervention, a way to have a memory of their experiences," Blackiston said. "I gave them a switch where they change color after they sense a certain light. What's nice about that is I can put them in a maze overnight and not watch them. The next day I can see what color they are and tell where they went in that maze."
This capability could prove useful for sensing heavy metals and other contaminants in bodies of water — or, in a more distant future, cancerous tumors in the human body. As demonstrated by the team's latest research, the xenobots can navigate a tube about a half-millimeter in diameter, raising the possibility of feeding them through an intravenous line, though Blackiston says the technology would have to be adapted for biomedical purposes.
Scaling up the technology is a major logistical hurdle. For now, Blackiston builds the xenobots by hand in a "fairly painful" lab-based process that involves putting cells together like Legos, he said. There are roughly 5,000 living cells in each of the xenobots. Peering through a microscope for hours on end, he uses microsurgery tools to remove unnecessary material and sculpt the xenobots into a spherical shape.
In the future, the team wants to use 3D bioprinting to continually produce thousands of the xenobots. A computational component of the process could also save researchers time.
"The computer, in the virtual world, can evolve the robots just like in the real world," Blackiston said. "It randomly moves cells around each generation, kind of like a parent, and you give the computer a goal. So you say, 'I want a robot that can walk in a straight line with four legs,' and it tries configurations of cells over millions of generations and it can solve that problem. I don't have to try all these different configurations and see what they do."
Whether the xenobots are true "robots" is a matter of debate among the researchers on the project. Though the xenobots don't fit the popular image of a mechanoid that moves and acts like a human but is made from artificial material, they are made to work on behalf of humans under their own motivation.
That places the xenobots somewhere on the continuum between humans and robots, said Josh Bongard, a professor at the University of Vermont who leads the computer science side of the project. He's excited for the possibilities of robotic biohybrids, or combinations of artificial and synthetic materials.
"The components that make up a traditional robot — the battery, the sensor and the motor — are themselves dumb," he said. "They have no in-built intelligence. This is one of the reasons why producing useful and safe robots is so difficult. At the level of the robot itself, maybe it's somewhat intelligent and adaptive, but any one part has zero adaptive capability. That's not true in living robots."
Created with living material, xenobots aren't biological organisms that have an evolutionary history on Earth, and can't be found on the tree of life. They were designed in a simulator for a specific purpose.
"This was thought up by artificial intelligence, essentially," Blackiston said. "That makes it special. The computer is coming up with solutions for humans that nature hasn't. I think that's going to be very powerful moving forward. What else can a computer evolve?"
The study, "A cellular platform for the development of synthetic living machines," published March 31 in Science Robotics, was authored by Douglas Blackiston, Emma Lederer and Michael Levin, Allen Discovery Center at Tufts University, Massachusetts; Sam Kriegman and Joshua Bongard, Department of Computer Science, University of Vermont; Simon Garnier, Department of Biological Sciences, New Jersey Institute of Technology; and Michael Levin, Wyss Institute for Biologically Inspired Engineering, Harvard University.