The Biological Robots Are Coming! But Note They Have Been Here for ∼3.5 Billion Years

Abstract

GEN BiotechnologyVol. 1, No. 1 Views & NewsFree AccessThe Biological Robots Are Coming! But Note They Have Been Here for ∼3.5 Billion YearsWilliam C. RatcliffWilliam C. Ratcliff*Address correspondence to: William C. Ratcliff, School of Biological Sciences, Georgia Institute of Technology, Atlanta, GA 30332, USA, E-mail Address: william.ratcliff@biology.gatech.eduSchool of Biological Sciences, Georgia Institute of Technology, Atlanta, Georgia, USA.Search for more papers by this authorPublished Online:16 Feb 2022https://doi.org/10.1089/genbio.2022.29007.wraAboutSectionsPDF/EPUB ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions Back To Publication ShareShare onFacebookTwitterLinked InRedditEmail Researchers led by Michael Levin and Josh Bongard cause a stir by investigating multicellular xenobot assemblies that challenge our conventional idea of replication.Where is the line between a robot and an organism? Does such a line even exist? Definitions, even those that are fundamental, have long been problematic in biology. For example, we still argue about how to define species,1 organisms,2 Darwinian fitness,3 or even life itself.4 Don't believe me? Go on Twitter and defend any definition of the mentioned concepts as the One True Definition!It should come as little surprise that pioneering work in biological robotics is as controversial as it is exciting. Take for example the article published in December 2021 in the Proceedings of the National Academy of Sciences by Sam Kreigman and Douglas Blackiston at Tufts University and colleagues.5 This article, entitled “Kinematic self-replication in reconfigurable organisms,”5 is the third installment of the authors' “xenobots” series.In their prior study, this interdisciplinary team of roboticists, developmental biologists, and computer scientists (which makes up the Institute for Computationally Designed Organisms) showed how xenobots can be generated through the self-assembly of pluripotent Xenopus laevis stem cells and programmed through artificial intelligence (AI)-guided design for functions such as walking locomotion.6 Xenobots were next shown to be capable of navigating aqueous environments through ciliary swimming, repair damage, and exhibit novel properties as swarms.7 In the latest report, the Tufts team has extended this model system, showing that xenobots are capable of reproduction.Before describing the article further, it is worth emphasizing that xenobots are an example of what is known in the trade as weird biology. The cells that constitute them do not divide. The only way they can reproduce is by building new xenobots from isolated cells in their environment. In the recent PNAS article, the authors take advantage of this fact. Using AI, the researchers designed xenobots with a crescent shape that leads to them swimming in circles, effectively pushing progenitor cells sitting on the bottom of a culture dish into piles. These piles then self-assemble into new xenobots (Fig. 1). This is not reproduction in the normal vertical manner, yet it is unquestionably reproduction.FIG. 1. Crescent-shaped xenobots swim in circles, sweeping up piles of isolated cells, which then self-assemble into a new xenobot. (Credit: Douglas Blackiston and Samuel Kriegman, Institute for Computationally Designed Organisms.)Such “kinetic” reproduction has its limits. For one thing, the offspring that are generated do not inherit their parents' ability to reproduce. Unlike normal vertical reproduction, where phenotype can be inherited by the transfer of informational polymers (i.e., DNA or RNA) or epigenetic cellular states, there is no material transfer between generations. Offspring xenobots are little more than piles of cells that self-assemble into spherical propagules, and thus lack the crescent shape that allows them to reproduce by swimming in circles and sweeping up piles of cells. I expect this is only a temporary limitation, surmountable by future innovations that allow the traits of daughter propagules to be tuned, perhaps by modifying the types of progenitor cells present and their rules for multicellular self-assembly.X, RobotThis study has generated much excitement in the press and a fair amount of criticism. Although it is easy to dismiss the apocalyptic “Terminator” media takes (even as the authors' YouTube video racks up hundreds of thousands of hits), more serious criticism has been lodged by the developmental biology community. First, many observers on Twitter thought this study was simply a rebranding of “animal caps,” a widely used model system in developmental biology that allows researchers to isolate pluripotent Xenopus cells and induce them to differentiate into different cell types.8 Second, individuals argued that calling xenobots “robots” was disingenuous, and that by the same definition fruit flies, yeast, and prions should be considered robots too.These criticisms miss the mark. Kreigman, Blackiston et al. are not attempting to learn about Xenopus development or claim that they are the first to show that animal caps can exhibit behaviors like swimming. The goal of the study is fundamentally robotic, exploring the rational design of biological machines capable of performing user-directed behaviors. As for the second critique, we come back to a problem of definitions. What is a robot and how does it differ from an organism?This distinction has always been blurry. In his influential 1976 book The Selfish Gene, Richard Dawkins wrote: “We are survival machines—robot vehicles blindly programmed to preserve the selfish molecules known as genes.” Whether you agree with this sentiment, all organisms possess the key trait of a robot: the ability to be programmed to perform autonomous actions. Perhaps the largest difference between biorobots and bog-standard organisms is that the former have behaviors programmed by people, the latter by a process of Darwinian evolution (and in organisms with cognition and memory, with the complex menagerie of behaviors developed over the individual's lifetime). According to this definition, all organisms are robots, although not all robots are organisms.The field of biological robotics is clearly in its infancy. Although I think it is reasonable to ask, “What practical use is there for xenobots?” in truth I do not think they have any immediate utility. The real payoff is the future potential of this approach. Bespoke multicellular construction is a key goal for synthetic biology,9 and currently it is far easier to leverage extant developmental modules (such as Xenopus stem cells) than construct them from scratch.Although the lack of multicellular heredity may seem like a crucial limitation, it can also be extremely valuable. Darwinian systems are inherently dangerous: difficult to control even under the best of circumstances. There may be real value in designing biological machines capable of performing a specific function, and then expiring, with little risk of gaining evolutionary autonomy.