They gather hundreds of freely moving cells together and then assemble the 'next generation' in a shape resembling the iconic 'Pac-Man' character's 'mouth.'
Do robots necessarily have to be made of materials like metal, plastic, wood, or concrete? Last year, researchers from the University of Vermont and Tufts University provided a negative answer to this question. They employed evolutionary algorithms to create the world’s first living robots, utilizing frog skin cells and muscle cells. These bioengineered entities, named ‘Xenobots,’ challenged the conventional notion that robots must be composed of traditional materials.
The ‘Xenobot’ stands apart from conventional robots and is not associated with any known animal species; rather, it is a novel, living, and programmable organism. Additionally, these entities exhibit autonomous movement and possess the remarkable ability to self-heal even when cut.
Xenobots can coordinate collective actions, such as moving in a circular motion (rotation).
Xenobot can exert force to propel external objects.
Xenobot can autonomously repair itself even after being cut open.
The robots exhibit a hollow structure in some designs, implying their capability to carry objects (such as drugs) to specified areas, holding significant research value and potential in fields such as medicine, biology, and chemistry. The related research was published in last year’s Proceedings of the National Academy of Sciences (PNAS).
However, a drawback in the initial version of Xenobot was its inability to achieve self-replication, setting the stage for subsequent research challenges.
In the latest issue of PNAS, the same research team announced that they have overcome this hurdle, creating the first-ever batch of living robots capable of self-replication.
The team found that these bioentities, designed by computers and manually assembled, could navigate to their small dish, locate hundreds of individual cells, aggregate them, and then assemble the ‘next generation’ Xenobot in a ‘Pac-Man’ shaped ‘mouth.’
In a matter of days, these ‘next-generation’ entities would assume the appearance and actions identical to their predecessors, continuing the cycle of seeking cells, establishing their copies, and repeating the process.
Xenobot is the world’s first batch of artificial intelligence-designed bio-robots capable of self-repair and self-replication. ‘With the right design, they will spontaneously self-replicate,’ stated Joshua Bongard, one of the leaders of the research and a computer scientist and robotics expert at the University of Vermont.
This groundbreaking research was published on November 29, 2021, in the Proceedings of the National Academy of Sciences.
Venturing into the Unknown
The ability for Xenobots to self-replicate was initially conceptualized by an AI program running on a supercomputer at the University of Vermont. Researchers employed an evolutionary algorithm capable of testing billions of biological body plans in simulations, with the aim of discovering cell configurations conducive to self-replication.
In the end, the AI identified a successful design: a group of cells resembling the iconic characters from 1980s arcade game Pac-Man.
Douglas Blackiston, a co-author of the study and a senior scientist at Tufts University, carved out the Xenobots’ body using the design provided by AI, utilizing a miniaturized soldering iron and surgical forceps. The body, comprised of 3000 frog cells, could move within a petri dish. Subsequently, frog cells added to the dish provided the building materials for the Xenobots’ body, which used these materials to craft Xenobabies in the ‘Pac-Man’ shaped ‘mouth.’ After a few days, the Xenobabies grew into new Xenobots bodies. By continually adding frog cell ingredients to the petri dish, this self-replication process could continue for generations.
In African clawed frog embryos, these cells would typically develop into skin. ‘They would be on the outside of the tadpole, blocking pathogens and redistributing mucus,’ explained Michael Levin, a biology professor and director at Tufts University’s Allen Discovery Center, and a co-lead of the new research. ‘But we put them in a new environment, giving them the opportunity to reimagine their multicellularity.’
The outcome of this reimagining was vastly different from skin. ‘For a long time, humans thought they had found all the ways life reproduces or replicates, but this way was never observed before,’ said Douglas Blackiston.
The cells have the frog’s genome, but they don’t transform into tadpoles. Instead, they utilize their collective intelligence and inherent plasticity to achieve astonishing feats,” stated Sam Kriegman, the lead author of the study and a recent Ph.D. graduate. In the early experiments, scientists were amazed that Xenobots could execute simple tasks as designed. Now, their surprise extends to the spontaneous replication of these computer-designed organisms. “We have the complete, unaltered frog genome,” said Levin, “but we didn’t deduce from it that these cells could collectively accomplish this new task (aggregating free cells and replicating the next generation).
“The replication process of these frog cells is vastly different from the way cells replicate inside the frog’s body. Any known animal or plant doesn’t replicate in this manner,” remarked Sam Kriegman, the primary author of the new study and a recent Ph.D. graduate.
The Xenobot, comprised of approximately 3000 cells, forms a spherical structure on its own. “They can reproduce, but afterward, the system typically collapses. In reality, sustaining continuous reproduction is extremely challenging,” Kriegman explained. However, with the aid of an AI program running on a supercomputer cluster, evolutionary algorithms can test billions of body shapes in a simulated environment, such as triangles, squares, pyramids, and starfish, to identify more efficient cells for motion-based replication.
“We discovered a previously unknown space within living organisms or life systems, a vast space,” said Bongard, a professor in the College of Engineering and Mathematical Sciences at the University of Vermont. “How do we explore that space? We found walking Xenobots, swimming Xenobots. In this study, we found Xenobots capable of self-replication. What’s next?” Perhaps, as scientists wrote in the Proceedings of the National Academy of Sciences, “Life beneath the surface harbors surprising behaviors, awaiting discovery.”
As shown in the above diagram, the replication process involves:
Isolation of stem cells from early frog embryos, dissociating them, and placing them in a saline solution (A):** Here, they aggregate into a sphere containing approximately 3000 cells. The sphere develops cilia on its outer surface after 3 days.
Placement of the generated group of mature cells among approximately 60,000 isolated stem cells in a 60mm diameter dish (B):** Their collective movement pushes some cells together into a heap (C and D). If this heap is large enough (at least 50 cells), it can develop into a mobile ciliated offspring (E). Additional offspring can be established by providing extra isolated stem cells (F).
In summary, ancestors (p) construct descendants (o), and then the descendants become ancestors. This process can be interrupted by subtracting additional dissociated cells.
Under currently known environmental conditions, the system naturally undergoes at most two rounds of self-replication. The probability of stopping (α) or replicating (1 − α) depends on the temperature range suitable for frog embryos, the concentration of dissociated cells, the number of mature organisms, random behaviors, solution viscosity, dish surface, and contamination probability. (Scale bar 500μm).
Opportunities and Risks
Compared to other known forms of biological reproduction, kinematic self-replication allows for significant enlargement and reduction in the size of each generation of offspring. This suggests that organisms might learn to autonomously design to produce offspring of different sizes, shapes, and useful behaviors, rather than just self-replicating in a numerical sense.
Some may find this research exciting, while others might feel concerned or even fearful about the concept of biological self-replication technology. However, for the scientific team, the next goal is a deeper understanding.
“We are working hard to understand this property: replication. The world and technology are changing rapidly, and it is crucial for society as a whole to study and understand how it operates,” says Bongard.
A supporter of Xenobots states: “This is a crossroads between robotics and biology.”
When asked if Xenobots are “intelligent,” Blackiston prefers to call them programmable organisms, with intelligence occurring in the design and programming stages rather than in the actual Xenobot. “My view is that they are not intelligent,” says Blackiston. However, he also acknowledges that this work challenges scientific definitions. “Due to these technologies, definitions are heading towards extinction,” says Bongard. “Xenobots are a product of AI, and AI itself is helping humanity eliminate the standard definition of intelligence.”
Facing numerous global challenges today, kinematic self-replication provides a means to deploy small-scale biological technologies, designed by AI replicators, enabling the replication process to be maximally controllable. Even though the performance of reconfigurable organisms is currently rudimentary, methods designed by artificial intelligence have proven to guide them towards more useful forms in the future.