Back when I was in high school learning basic Cell biology, I recall being baffled by one mystery. We were being taught about Cells in great detail. Their internal components and processes. I recall things such as protoplasm, cell membranes, the nucleus, mitochondria, as well as DNA, RNA and the ATP cycle. All very well for understanding the operation of a single cell, I recall thinking. But how does that explain the operation of the whole body? It’s various complex organs and their complicated interactions. Even the development of the embryonic cells into the organs and tissues of the multicellular organism? What mechanisms make those cells understand that they are part of one organ and the organs understand that they are part of one body? Surely, I felt, there must exist some mechanism that tells each cell what it must be, what it must do, some signals that make it do completely different things from other cells in the same body. DNA could not explain this to me, wasn’t the DNA the same in each cell? Nor could it be the brain. The brain doesn’t even exist in the early embryo!
Try as I might, I could not find a satisfactory explanation. Granted, it was the pre-internet era, and my quest was confined to asking teachers, senior students and checking the books I could find in the library. The only answer I was given was that it was hormones. I was told, chemical hormones in the body conveyed to the cells what they must do. That was neither convincing nor satisfying to say the least. But I let it go, thinking someday I will find a more satisfying answer.
The question remained unanswered the next 35+ years. Having graduated with a master’s in computer engineering, followed by a 30-year career as a computer engineer, far removed from biology I did not have reasons to keep up with advances in Biology. And out of nowhere, this year in 2023, I came across some fascinating talks by the brilliant Dr Michael Levin, distinguished professor at Tufts University. Finally, there it was, a satisfactory explanation for multi-cellular organization, an answer, coincidentally remarkably relatable to my day job. As I found myself getting pulled into the orb of Dr Levin’s research and his fascinating, persuasive and thought provoking ideas at the frontiers of science, I developed this irresistible urge to share these insights along with the deep questions these ideas evoke. If you have ever wondered how a single cell (fertilized egg) knows how to develop into a fully formed baby, read on to be amazed by many other unimaginable feats cells perform.
Dr Levin, perhaps not unexpectedly, is a computer engineer turned biologist who was obsessed with these same questions. Dr Levin’s groundbreaking discoveries and out of the box thinking may not only create exciting new therapies for cancer and other degenerative diseases but also be the impetus for a significant leap forward in our understanding of Biology and perhaps a profound new perspective on life itself.
In this article I will first review the new insights and claims, before taking you down the rabbit hole of bewildering immortal beings and synthetic bio-bots and end with a speculative cogitation on the broad impact of the new biology that will inevitably emerge before our eyes.
Surprising New Insights
Dr Michael Levin’s experiments have led to surprising new insights and claims that challenge conventional views and understanding of cells and their behavior. The key insights may be summarized as follows:
All cells are competent cognitive agents capable of problem solving to reach goals.
The conventional gene-centric view has not really granted each cell its own agency. They have been treated as merely mindless automatons faithfully executing specific functions they were programmed to do by the genes. Nor is there any notion of cell level goals in the conventional view.
Multiple cells can connect with each other to form collectives that reach for bigger goals.
The conventional view has no notion of cell level goals and consequently no theory for the goals of a cell collective. In the conventional view, goals, purpose etc are higher order emergent phenomena confined to brains. Whereas in Dr Levin’s findings there is a hierarchy of goals that get bigger as you scale up the cell collectives. From the limited goals of a single cell to goals of a tissue, to goals of an organ and organism as a whole. Dr Levin terms this as Multi-scale Competency Architecture, because at each level the cell collective can solve problems it encounters while growing itself or maintaining its function in-spite of perturbations or disruptions.
Learning and Memory is a capability common to all cells and not just neurons and brains.
In the conventional view we tend to think of the ability to learn and retain memories as exclusive to complex animals with brains. In reality, there are numerous examples that clearly show this to be false. In fact, Levin lab’s work shows that neurons and brains are just one special case of learning and memory for solving problems in a 3D space and time domain, but other cells may have the ability to solve problems in much higher dimensional spaces. In the case of memory there is growing evidence that memory may not be simply in the synapses of the neurons but distributed in other ways across the tissues of the body.
DNA is not a blueprint for the entire organism. DNA is only a blueprint for the machinery (hardware) of each cell.
The conventional gene-centric view assumes that DNA is the blueprint for the entire form and function of the organism. A necessary and sufficient encoding of all the processes involved in building and maintaining the organism. But having the ability now to read genomes we can clearly see that the genes mostly encode the sequences for creating proteins, these are the hardware building blocks of cells. Thus the DNA and genes contained therein can be thought of as plans for assembling the hardware components, the proteins, required to grow and maintain the cells. Thus the DNA does not contain the program i.e., the software algorithm (continuing the computer analogy) per se that directs which proteins to synthesize when, the so called gene expression process. The software program must necessarily be encoded elsewhere.
Memory of body form is encoded outside of DNA as a goal state that acts as a set-point that guides the development of the body during embryogenesis or tissue development during limb or organ regeneration.
The conventional view assumes that an organism’s form is somehow encoded in the DNA or arises as emergent complexity of self-organization in the local interactions of the cells. The cells are thought to be following simple local rules at the chemical level. Dr Levin’s lab has experimentally shown that the body plan is encoded as bioelectric voltage on the surfaces of the cells. Levin lab has demonstrated how to read the body plan. Encoded as a pattern of dots (electric voltage level at each dot) spread across the embryo cell collective. The pattern so encoded can be thought of as a set-point describing the final goal state, a description of the anatomy. Compared to the temperature set-point of a thermostat, anatomical set-point is a much more complex description because an embryo is a more complex system. A thermostat is a simple control system loop that adjusts a heating element up/down until the set temperature is reached. In an embryo, the bioelectric dot pattern guides the cells during development when they are self-organizing the body. Each cell taking its cue from the voltage level on its surface as to the actions it must perform, and executing the control loop that measures the distance to the set-point and performing actions that continuously reduce it.
The memory of body form can be altered (edited) to set a different goal state to make the tissue develop into a different body shape without requiring any changes in genes or DNA.
This would be near impossible in the conventional view, because the conventional view is that genes and DNA control everything and knowing which genes out of 25000 or so, to manipulate to get the desired body or limb shape is a combinatorial nightmare to predict. On the other hand, Dr Levin’s lab has shown how the dot pattern of the body plan could be edited relatively easily to generate novel yet functional body shapes or to help regenerate lost limbs in animals that don’t have that innate ability to regenerate them.
Cancer is the result of loss of cohesion of individual cells with the collective goal state of their tissue.
The conventional view correctly identifies specific cancer-causing genetic defects but fails to fully explain the behavior of cancer cells such as metastasis — the spread of cancer from few cells in one tissue to many unrelated tissues throughout the body. The cognitive agent model of Cells begins to further our understanding of how this could occur and potential ways to reverse it.
The novel insights from Dr Levin’s work has profound implications for regenerative medicine. The fact that we can interact with cells at the higher level of goal states and leverage their own intelligence to carry out self-repair, promises to be very powerful. Much more so than the much celebrated leading edge techniques such as CRISPR, editing genes at the molecular level. This is because same genes may have multiple effects and there is no way to predict what other processes are affected by changing some genes, creating side-effects and unintended consequences.
Have medicine and biology have been too reductionist in their approach? Understanding details of molecular pathways involved in certain cellular function may be very useful, but trying to work out what higher level goals are being pursued from the bottom up may be a fools errand. A top down approach that understands the higher level controls the cells employ may be much more effective for medicine. The conventional molecular medicine approach in a crude analogy is akin to trying to launch a Smartphone App by opening the phone case and fiddling with internal hardware instead of tapping the icon on the screen. The new approach led by Levin lab may be far more effective and simpler, just figuring out the icons and gestures that the Smartphone user interface presents to launch and control the Apps.
Intelligence & Cognition
Since much of the insights out of this research revolve around the notion of intelligent agents it is necessary to be precise about the definition of Intelligence and Cognition. The definition adopted by Dr Levin is apt, taken from the work of philosopher and psychologist William James, it states simply that ‘intelligence is the ability to achieve the same goal by different means’.
In this view reaching goals can be thought of as a navigation or search problem. A goal may be described as a point location in a particular multi-dimensional space and intelligence is then defined as the ability to find a path from some other point in that space to the goal.
Humans generally acknowledge intelligence in other creatures if they observe a creature carry out such a navigation through obstacles in ordinary 3D space, imagine a rat in a maze finding a way to some food treat. But there are many other unconventional spaces where humans and cells carry out such problem solving. As an example of an unconventional space, imagine a cook preparing a meal. The cook is traversing a path in a multi-dimensional space consisting of dimensions such as flavor, aroma, color, temperature etc to reach the goal point of perfectly-delicious. Adjusting each parameter just right along the way.
Likewise, the spaces cells operate in are diverse. Embryogenesis operates in what Dr Levin terms as Morphospace. A space of formed by the parameters involved in defining an organism’s shape, size and symmetry. We call the navigation performed by embryonic cells in this morphospace as problem solving because the embryos reach the morphological goal (the final shape and anatomy of the organism) regardless of the starting point configuration. There are other spaces such as transcriptional and physiological in which cells operate and solve amazingly complex problems. In the following sections we will look at some astonishing examples of such problem solving in the morphological realm.
Astonishing Cellular Intelligence
The popular perception among the scientifically literate is that Cells contain genetic code in the form of DNA and DNA encodes all the necessary information for the development and functioning of the organism. Evolution is understood to be the effect of natural selection on chance mutations of this genetic code. Evolution passes on to the next generation the genes that confer better fitness. In this view it is hard to consider Cells as having intelligence of their own. We tend to think of Cells mechanistically following some software like program encoded in the genes to go about the business of development of the body and its ongoing functioning through metabolism. But when looked at closely, cells exhibit some curious behaviors that are hard to explain as mechanistic.
Morphospace Intelligence
We often refer to the development of a complex organism like a human being from a single-cell zygote (union of a sperm and egg) into a fully formed baby as the miracle of birth. It appears to be a reliable process that produces the correct internal organs and external form including the body symmetry, the correct number of limbs and organs, with only minor variations in size, skin tone, eye color etc. The process appears almost pre-programmed to reliably produce the same form. But there are clues in some observations that there is something more than meets the eye.
Human Embryo
If you were to perturb the early embryo in various ways it does not destroy the embryo, surprisingly it appears to reach its final form (the goal) regardless. For example, if you cut some early embryos into half, both halves go on to make identical twins. Each half of the embryo develops to the same end state of morphology.
Picasso Tadpoles
In one experiment performed at the Levin lab, a tadpole’s anatomy was scrambled to move it’s parts to weird locations, mouth on the side of the head, the eyes in the back etc, like a Picasso painting. Astonishingly, during metamorphosis from tadpole to frog the parts rearranged themselves correctly to produce a normal looking frog. When this process is observed in real-time what you see is that the parts move around and keep adjusting to find their correct location. Sometimes overshooting and then adjusting back. The idea of some sort of morphological set-point guiding this process, as a thermostat would, seems self-evident.
Axolotl Salamander Regeneration
Some amazing animals like the Axolotl Salamander (endangered and endemic to Mexico City) are noteworthy because unlike other amphibians they do not undergo metamorphosis (development from tadpole like stage to adult frog). This fact probably endows Axolotls with the ability to regenerate eyes, limbs, jaws etc., forever into adulthood. You can cut a limb anywhere and it regrows that limb and the process automatically stops when the limb has grown back to the perfect proportion. How do the cells of the limb know when to stop rebuilding? Again, some sort of set-point appears to be at play.
Newt Kidney Tubules
One of the most interesting examples of morphospace problem solving in the face of perturbations is known from the 1940s. In this experiment the embryonic cells of a Newt are made bigger in size by preventing their division for some time during development. What is observed when this is done is that anatomical parts of the Newt do not become bigger in spite of the fact that the cells are bigger. In the figure below are illustrated the kidney tubules in cross section. When the cells are of a normal size, the tubule is formed by arranging 8–10 cells in a ring and leaving an opening in the middle. When the cells are bigger in size the embryonic development automatically adjusts to fewer cells in the ring to make the same size ring. And most amazingly, if the cells are made truly ginormous, then each single cell of the tubule bends itself around in a ring to form the tubule of the exact same size. That implies that the same anatomical goal of tubule shape and size is achieved through entirely different means of construction.
The examples above indicate that the standard notion of development and regeneration as being an open-loop process starting with genes expressing proteins and cells following some local rules leading to emergent complex forms is unlikely to be true. If you have ever seen the Conway Game of Life, where simple rules for cells in a grid pattern updating their state based on states of their neighbors (Cellular Automata in general) you know that simple rules can create endless complexity. This is where the conventional view of emergent form comes from.
But from examples above, what Dr Levin is pointing out is that there isn’t just local rules that is affecting the outcome, there is evidently some feedback about how far the current state is from the goal that is guiding the process in each cell. This feedback indicates some sort of a closed loop process, much like the thermostat model. Something is measuring the error or deviation from the set-point and applying the corrective inputs to guide the process towards the set-point. A very similar error minimization process is employed in Artificial Intelligence during neural network training.
Set-points and related mechanisms are well known in biology, the process is known as homeostasis. But this anatomical homeostasis differs from typical physio-chemical homeostasis, instead of the set-point being some one dimensional quantity like blood pH, the set-point in this anatomical context is a complex multi-dimensional state space matrix. How is that set-point encoded in the cells? Is there some sort of map encoded in the cells that the cells check their progress against? Dr Levin theorized that some map like this must exist and if it can be found and edited one could theoretically control the process of morphogenesis and regeneration at the higher level of the map rather than messing at the low level of genes and molecular pathways. Amazingly, they not only found the encoding of the map, they were able to edit the map and change the regenerative outcomes to prove the theory. The medium the cells use to store these maps as memory is called Bioelectricity.
Bioelectricity
It is well known that brain cells or neurons use electricity to store and process information. The hardware of neurons is composed of small protein channels called ion channels that maintain a voltage across the cell membrane and also propagate it to their neighbors via synapses using gap junctions. What is less well known is that these ion channels and gap junctions are not exclusive to neurons. All cells have ion channels and communicate through gap junctions with their neighbors. They do not have the dendrites and long axons like the neurons, but they do exchange electrical and chemical signals with neighboring cells by aligning the gap junctions on their surface. The only difference is that the speed of this signaling is slower in the regular cells compared to the signaling speed in the neurons. This is not surprising, neurons or nerve cells did not magically appear with this bioelectricity, the evolutionary precursor of neurons must have been regular cells that first developed the bioelectric capabilities, eventually enhanced and specialized by evolution into the neurons with high speed processing capability. The figure below depicts the ion channels and gap junctions for various ions such as Sodium (Na+), Potassium (K+), Calcium(Ca++) and Chlorine (Cl-).
Bioelectricity in Embryogenesis
To study what bioelectricity is doing in frog embryos, Dr Levin’s lab borrowed techniques from neuroscientists who have been studying electrical activity in the brain for a long time. Newly developed techniques using voltage sensitive dyes that light up cells on electrical activity have been used recently to study brain activity in living animals. The same dyes were then used to watch developing embryo cells communicating in real time.
The image above shows a frog embryo lighting up with electricity as the cells of the embryo exchange information during development. The colors indicate the voltage levels and clearly shows how the pattern of electric voltage changes slowly over time (matter of hours, as opposed to brain activity patterns that change in milliseconds). In the brain the electrical activity controls muscles of the body to effect motion of the organism in 3D space. Likewise, one can theorize the slower electrical activity in the embryo is effecting a movement of the cells in morphospace. Movement in morphospace not only implies motion in 3D space, but also cell replication, differentiation, and the control over shape and size.
Interpreting the Bioelectric Patterns
In some cases the bioelectric patterns that constitute the map telling the embryo what to build appear quite obvious. In the figure below the top two grayscale images show the pattern formed on the surface of the embryo just prior to the development of the tadpole face. In this case we can clearly make out the pattern is indicating where the eyes and mouth need to be. Most patterns are not so obviously correlated to the eventual morphology.
The bottom images in the figure show what happens when some cancer causing genes (oncogenes) are deliberately injected into some part of the tadpole. Even before the tumor starts forming the bioelectric pattern indicates the cells with the oncogenes are changing their voltage and disconnecting from the cells around them. Once disconnected, those cells treat the rest of the cells as external environment and revert to bacteria like behavior and start multiplying indiscriminately to form the tumor.
Editing Bioelectric Patterns
Using tools developed by neuroscientists to study neural electric circuits such as Optogenetics and certain drugs, Dr Levin’s lab was able to manipulate the bioelectric patterns. These techniques can add ion channels and turn them on/off to create different patterns. In the figure below we can see one such experiment. By studying the bioelectric patterns for the tadpole eye in the wild type (WT) they were able to successfully recreate that pattern of bioelectric voltages in an entirely different group of cells, the cells that would normally form the gut of the tadpole. To their amazement the gut cells responded to the edited bioelectric pattern and faithfully started forming the eye at the ventral gut area. Not only that, when the edited cells were too few to form the eye, it was observed that those cells recruited nearby non-voltage modified cells into helping them form the eye in the wrong place (ectopic). Eventually, the misplaced eye becomes fully functional by forming neural connections to the spinal cord.
Limb Regeneration
The Levin lab has used the bioelectric pattern modification technique not just to reshape embryonic cells, but also to induce regeneration of amputated limbs in animals that do not naturally have that ability. The figure below illustrates successful regeneration of a frog hind leg using a cocktail of drugs that reprograms the bioelectric pattern at the wound site where the leg was amputated. Then protecting that limb for several weeks as the regeneration progresses. Normally, a wound would simply heal, leaving a scar, but in this case the cells at the wound site were induced to regrow the entire limb, back to complete sensation and mobility. This begs the question why don’t all organisms have this regeneration ability by default? Why do we form scars instead of complete regeneration? Nobody knows for sure, but it may be speculated that because regeneration requires some special care of the tissues for several weeks it might be evolutionarily more expedient from a survival standpoint for these organisms to quickly heal with scar tissue and move on.
Imperative vs Procedural Programming
Normal computer programs are said to be Procedural. Each program codes a set of procedures detailing every little step that a computer must perform to carry out a computation or algorithm. For example, to move an icon on a computer or phone screen, a computer program would detail the set of steps to draw the icon in each new position and erase it from the previous position.
In contrast to this procedural paradigm, a different programming paradigm is now employed by the large data center infrastructure of Clouds. Called Imperative programming. These clouds contain thousands of computer servers, and the sheer number implies that some hardware failures and outages are occurring in the network at some frequency. To mitigate these failures the network is designed to automatically reconfigure itself to continue operation even when multiple failures may happen in quick succession. But how to reconfigure the network is not programmed procedurally into the servers. The imperative programming technique simply spells out the end state of what a healthy network looks like. The servers themselves figure out how far the current state is from the goal state and adjust the work among the survivors to keep the service running.
As in the Cloud data centers, the bioelectric patterns used by biology seem to be an imperative programming model for the regeneration of body parts. The pattern does not encode any detailed procedures for rebuilding missing parts. The cells are smart enough to figure out how to do that on their own. They simply need an indicator for the goal state to be sought. We programmers and engineers are simply mimicking the techniques evolution perfected millions of years ago.
Planaria
The one animal that may be completely upending our perspective on life is a small two inch worm called Planaria. These amazing worms are the champions of regeneration. You can cut them into hundreds of little pieces and each piece regenerates into a perfect new worm. As illustrated in the figure below any slice of the worm automatically figures out what parts are missing and recreates the remaining to form an entire new worm. This is not entirely surprising because these worms do not reproduce sexually, they reproduce by doing exactly this, tearing themselves into two pieces.
The other curious facts about Planaria is that they do not age and are immortal and cancer free. If you look at their genome it may appear horrendous, because they accumulate good and bad mutations over eons of such reproduction. But the ability to regenerate faithfully is preserved regardless of the contents of the genome that may have particularly bad mutations.
Two Headed Planaria
In another illuminating experiment with Planaria, the Levin lab was able to identify the bioelectric pattern that controls the regeneration and then edit the pattern such that the regeneration produced a different result. In the figure below the images on the left show the pattern of a normal worm. You can see the voltage map indicates where the head is supposed to be and where the tail is supposed to go. The images on the right show the edited pattern that is indicating that there should be heads on both ends, no tail. After this editing, when the worm is sliced, the sliced portion creates a two headed worm — heads on both ends. It is important to note that the pattern was edited in the body of a one headed worm, but the two heads resulted after slicing that one headed worm.
This means that the pattern is a memory of what to do in the future in the event of an injury. It is not a pattern showing what the current shape of the organism is.
This pattern memory is not encoded in the genes. No genes were edited and yet the memory lives on in successive generations. You can keep slicing and dicing successively and each generation produces two headed worms. So morphological shape or phenotype of the organism is determined entirely by the bioelectric pattern memory.
The astute reader may wonder how the bioelectric pattern encoded across the whole worm can be remembered by each slice of the worm? No matter how many slices we make? Nobody knows for sure, but it is speculated that the patterns may be holographically encoded in the cells. Holograms are not like regular photographs that have a flat 2D representation of light on a photographic medium like film or digital camera pixels. Holography records an interference pattern of lasers illuminating an object onto the photographic medium. In a normal photograph, if you cut it into small pieces, each piece only contains the fraction of the object image that was recorded on that piece. In a Hologram in contrast, you can take a small slice of the Hologram and still be able to see the entire object. Perhaps the bioelectric pattern encoding is like a holographic image rather than a like a photographic image.
Planaria in Barium
In another unrelated but mind blowing experiment with Planaria Dr Levin’s team put the worms in a Barium solution, which is highly toxic to the Planaria. It literally explodes the heads of these worms. And yet, after some time the headless worms regrew new heads and the new heads were now adapted to a Barium environment. When the Levin lab team analyzed the genome to see what had changed, they discovered the worm tissue had mutated a handful of genes that made it much better adapted to Barium. This mutation appeared to persist for a few generations. Eventually keeping the worms in normal Barium-free environment the subsequent generations lost the Barium resistance. Clearly, the cellular intelligence was solving problems beyond the morphological space in some transcriptional space, choosing the right genes to modify to become Barium resistant.
Agential Behavior
To understand how far the cognitive agent behavior and plasticity can be pushed Dr Levin and colleagues asked the simple question: what happens when the cells are taken out of their context and left to their own devices? Do they do something different? Do they die?
Xenobots
To answer these questions they took the cap cells (cells fated to form the skin) from a Xenopus frog embryo and put them in a petri dish. The figure below depicts what actually happens. The cells do not die, but they gather themselves into a ball that starts to show its own behavior. Dr Levin and colleagues have named these ball like creations as Xenobots, combination of Xenopus and Robot.
Xenobot Behavior
The first thing Xenobots do is to start using the little hair like fibers called cilia that are present on the surface of these skin cells in a completely novel way. The cilia are normally used by the skin to spread mucus on the surface of the skin. That protects the skin from parasites. But in the Xenobots the cilia are used as little oars to row the Xenobot in water. So these Xenobots quickly figure out how to swim around and navigate space. The figure below depicts the Xenobot traversing a maze. Even making spontaneous decisions like suddenly reversing direction of motion. How does a pile of skin cells have this behavior? There is no brain, neurons or anything like that present in these xenobots.
Xenobot Replication
The most intriguing behavior was observed when Xenobots were given a loose collection of additional skin cells. In the figure below the image on the left shows the loose skins cells (white powder like substance) and a collection of orange colored Xenobots. What the Xenobots do when they discover the loose cells is to gather those loose cells into piles and keep shaping those piles until those piles become a new set of Xenobots.
This is Xenobots self replicating by finding other material in the environment. What is termed as Kinematic Replication. Of course, the loose material is itself Agential material. Material that is quite clearly competent in its own right.
Anthrobots
As recently as November 2023 the Levin lab has published a new paper detailing the successful creation of bots from Human trachea cells, christened Anthrobots. Remarkably, these Anthrobots are created not from human embryonic cells, but from adult cells. And yet, when liberated from their normal environment they assume novel behavioral patterns not too different from Xenobots.
Implications
Beyond the tremendous potential benefits to regenerative medicine, such as the ability to correct birth defects, cure cancer and other degenerative diseases, the results of this research calls into question so much of what we have hitherto understood about biology, evolution, and life.
Upending Paradigms & Dogmas
I see at least two paradigms being upended by this research. First, is the Gene-centric view of Darwinian Evolution, and second is the Neuro-centric view of cognition.
Gene-Centric Evolution
The gene-centric view of evolution is the current dogma of Biology. Most well explicated by biologist and author Richard Dawkins, who is famous for introducing the concepts of the Selfish Gene and the term Meme. In this view, natural selection operates on genes, and the gene is the central character trying to propagate itself, and the organism’s phenotype is simply a vehicle to that end. Phenotype are the attributes such as morphology that define the species of organism. In the gene-centric view phenotype is endowed to the organism by the genes. But as we have seen in the examples above, genes are only one factor in the phenotypic morphology of the organism. The organism has a lot of plasticity in its morphological shape, that shape is not rigidly tied to the genes. And the shape itself is encoded outside the genome. Moreover the material shaped is itself a competent cognitive entity with malleable boundaries. The story of evolution appears to be much more interesting than the Selfish Gene would have you believe i.e., a straightforward mapping from Genes to Phenotypes.
It appears that what evolution has evolved, are intelligent problem solving systems that operate at multiple scales. Genes appears to be mostly a hardware component in that system and the phenotype is the result of the problem solving ability of the collective intelligence of the cells with genes and pattern memory as inputs.
What Dr Levin is arguing is that evolution of multi-cellular organisms is not operating at the molecular level, where searching for the right combination of molecules that lead to better fitness of the individual would be enormously difficult. It is instead operating with cognitive material (cells) and shaping their behavior with appropriate signals to produce macroscopic effects. In Dr Levin’s view individual cells are agents with specific behavior, and on their surface they expose a set of ion channel interfaces, with which electrical or chemical signals can be sent and received. These signals can be thought of as the language used to coax the cell to modify its behavior. Any external agent that learns this language can control the cell or shape its behavior. Cell collectives can thus control the individual cells. Dr Levin compares this to computer hacking. In his view everything in Biology is hacking everything else by learning their language.
Evolution in this view is not searching by random mutation the combinatorially intractable space of gene combinations, but a much more tractable space of behavior shaping signals. Just as the Levin lab was able to create a two headed Planaria, evolution too could create one in short order without experimenting with different gene combinations over many generations.
In a recent debate that Richard Dawkins had with the British physiologist Denis Noble, Noble argued for such a revision of the gene-centric view. To a physiologist like Noble, the two decades since the human genome was first sequenced, have brought little in terms of the significant advances in medicine that were promised. Dr Levin’s approach to medicine appears much better aligned with the view supported by Noble. Noble compares genes to a musical instrument and the living cell to the process that breathes music into it.
The conventional gene-centric dogma has trouble accepting the agency of cells affecting evolutionary outcomes. This may be because there is a fear in the scientific community that crediting cells with agency and intelligence could be the slippery slope toward Intelligent Design, which smacks of Creationist ideology. But such fears are signs of insecurity. True scientific attitude must follow the evidence. It must attempt to account for where single cells acquired intelligence in the first place. If there is some natural, mechanistic explanation for how cells acquired cognition then the stigma of Intelligent Design could be easily shed.
Neuro-Centric Cognition
The second paradigm upended is surely the Neuro-centric view of cognition. Using Dr Levin’s definition of intelligence and cognition, it is clear that intelligence pervades all of Biology. Have we been too narrow in our definition of intelligence? Biased as we are towards recognizing intelligence only in the familiar problem domains and explaining away or even ignoring remarkable intelligent behaviors as some sort of innate instinct. What Dr Levin’s research is highlighting is the diversity of problem domains where cells and cell collectives operate, and solve hard problems, in ways our brains can hardly begin to fathom. What appears to be common to all this intelligence, is the entity endowed with intelligence. We have been referring to that entity as an Agent. It is those Agents that appear to be solving problems and seeking goals. So what are these Agents anyway? Are Cells and Cell collectives the only things that are capable of being Agents? Let us now try to dig a bit further into this question.
What constitutes an Agent?
So far in this article we have referred to Cells and their collectives as Agents. Agents are physical entities that have beliefs, desires and competency in actions. In other words, Agents are a composite of minds and a physical boundary, a boundary with which they can sense the environment and effect action. Action internal to themselves as well as on their environment. I use the term mind because Agents are capable of making decisions about what action to take on some external stimulus, that decision is not determined purely by physics. But rather reached by some internal computation, making it hard for an external observer to predict.
Biologists have shunned attributing the notion of minds to anything other than complex brains, Michael Levin and philosopher Daniel Dennett have argued that this overcautious refusal to anthropomorphize behavior may be holding back Biology. This visceral rejection of anthropomorphism has led Cell Biology towards an extreme focus on the seemingly ‘mindless’ interactions of molecular pathways and ignoring the behavior of Cells at the higher level. Levels at which they appear to be entities with minds. Understanding Cell behavior at these higher levels is surely the faster way to create more effective medical interventions.
A key facet of Agents is their competency horizon, the things that they are able to sense, remember, anticipate, choose amongst and crucially, affect. Dr Levin terms this as an agent’s Cognitive Light Cone. The Agent’s light cone maybe characterized by the size of its goals in space and time, Anticipation implies that the Agent has a model of itself in relationship to its environment. When the Agent learns, it is updating its model for better prediction into the future. So the light cone is characterizing the space and time reach of the Agent. How far in space it has the competence to sense and affect, and how far back in time it can remember, and how far into the future it can predict.
In this view it is easy to apprehend that a single Cell has a cognitive light cone much more limited than of a large cell collective. A collective is uniting the competency of the whole swarm of cells. The swarm can perhaps sense a larger environment, remember more, anticipate more and affect more. Humans as Agents of course have extremely large light cones, goals extending far into space and time. How do Cells combine their smaller light cones into the larger light cone of the collective? It is by joining together and exchanging information, either through bioelectric or chemical signals.
When cells in a swarm are coupled tightly and exchanging such signals, it is likely that a cell is unable to distinguish which signals came from within it’s own boundary and which came from another cell. Imagine, the case where one cell on the boundary of the swarm reacted to some stressor in the environment and sent a chemical message to its neighbor, which passed the signal to its neighbor and so on. In this scenario the boundary of the self of each cell is erased and the entire swarm behaves as one larger collective self. The large collective self then acts as a larger Agent using the sensing abilities of the cells along the boundary, while at the same time it is able to multiply the force it can exert on its environment. Such factors go on to expand the competency horizon of the swarm. Whether some cells within the swarm act as leaders and are able to model the big-picture and persuade others in the swarm is an open question. There are certainly specialized cells that behave as instructors in some situations.
Self & Cancer
Looking at our bodies as a collection of cognitive agents cooperating as a coherent single Self, a self that has goals and that maintain itself leads to a new understanding of Cancer. Cancer does not appear to be a genetic disorder, but a disassociation of the larger self into smaller selves. The affected cells lose their identification with the larger self and their cognitive light cones shrink drastically. Both spatially and temporally. First, losing their bioelectric connectivity with their neighbors the cells can no longer identify their spatial boundary as the boundary of the swarm, and they regress to their own boundary of a single cell. Secondly, losing the access to swarm memory and it’s lifetime of learned behavior, the cells’ memory and goals revert to much shorter time spans.
Dr Levin’s research has demonstrated in one experiment that cancerous tumors can be induced in healthy tadpoles by blocking the bioelectric signals from certain instructor cells to melanocyte cells. In that case the melanocytes (who have been cut off from the instructor cells) go crazy, and start behaving like single cells, spreading melanoma all over the body. In this experiment nothing about the genome was changed. Purely messing with the electrical signaling produces this cancer like behavior.
Given this understanding opens up new possibilities for curing Cancer by reconnecting cells and restoring their sense of self. This has been demonstrated in other experiments. In these experiments, cells were deliberately injected with cancer causing genes, and where tumors would have certainly formed, they were successfully suppressed by injecting new ion channels that kept the bioelectrical communication going. Communication that would otherwise have been shut down by the cancer genes. The interesting thing is that well known FDA approved ion channel drugs were used to induce the new ion channels and keep the communication channels open. Not requiring any new drugs to be invented. This research underscores the need to develop a deeper understanding of these cell communication patterns and how to intervene with such electroceuticals to tackle diseases like Cancer in humans.
Closing Thoughts
How do we make sense of the implications of the complex agential behavior exhibited by cells? Surely, the self-similarity in agential behavior at every scale is implying some deep truths. If we are constituted of a hierarchy of intelligent agents at different scales, who are we really? Where does cognition begin? From single cells on up for sure, but does it also go deeper? Inside the cells? In the components of cells? How far down does it go? Also how far up does it go? Are we fully developed humans the top of the hierarchy or are we simply parts of an even higher level agent? Are our social networks an agent? How about the entire Biosphere? Is it operating as an agent? How about the solar system and the galaxy and the Universe? Is it turtles all the way down and up?
Dr Levin and collaborating scientists such as Chris Fields, Karl Friston and others are investigating such deep questions. Both from a philosophical standpoint and to make the research in Artificial Intelligence and Artificial Life more ethical and sensible. For example, a lot of money appears to be directed these days towards creating Artificial General Intelligence where the goal appears to be to mimic human intelligence. Surely, this appears to be misguided, now that we know that human intelligence is one small corner of the diversity of intelligences that exist and are possible. The smart money should be going towards creating AIs that complement human intelligence and not in trying to replicate it.
Scientists like Chris Fields and Karl Friston believe that intelligence pervades the universe and is perhaps not confined to biological agents. In their view intelligent agent like behavior goes all the way down to fundamental particles. In a new physics proposed by Chris Fields, the reductionist approach of particle physics that is trying to split particles into more fundamental units is doomed to failure and that all physics can only be viewed as interactions between agents or observers. In effect the reality of the universe or any system is unknowable except in terms of the effects observable to the observer. The attractive property of this kind of physics is that it is scale free. The same theory can be applied at the scales of fundamental particles and the scales of the cosmos depending on what scale you want to make predictions. This has clear echoes with Dr Levin’s work with nested biological agents.
If intelligence pervades the universe, I wonder, why we have not found any agents, (at least at the scale of Classical physics) other than biological life? Until we do, I think we must treat the biological life on earth as precious, and treat it with the respect and reverence it deserves. We, the biological beings on this planet are all descendants of a common ancestor. An organism theorized to be the single celled organism termed LUCA (Last Universal Common Ancestor). Our DNA and other cellular material is derived from billions of years of division and replication from that one original cell.
In one sense all beings on earth that have ever lived are thus related, separate individuals that were born, lived and died. But in another sense, I can’t help but wonder, what we call evolution of life on earth, isn’t actually one long drawn out process of morphogenesis? Starting from the single seed LUCA, could it be one inexorable process of exploration of the planet’s environment? Carried out by that single self spreading its tentacles over the globe. Pulsating in time as various different forms of agents that we call species. All cooperating in a living connected whole, a Self, we call the biosphere. Surviving billions of years, through hundreds of planet scale catastrophes. That self persevering through catastrophes, never snuffed out, even though many species of agents are gone for good. And a chill runs down my spine as I contemplate the unavoidable thought, of the failure mode in this living process, the creation of an agent, a species, that disconnects itself from this biosphere-self and regards the whole as an external environment. Is it a cancer? Will it snuff out the self that has persisted these billions of years? Or will it reconnect with the biosphere-self and restore itself? Learn to be One with the Whole?