New clues to the origins of biological intelligence
In the middle of his historical book About the origin of species, Darwin had a crisis of faith. In a fit of honesty, he wrote: “Suppose that the eye with all its inimitable devices for adjusting focus at different distances, for admitting different amounts of light, and for correcting spherical and chromatic aberrations, might have been formed by natural selection seems, I confess, absurd in the highest degree. While scientists are still working on the details of the eye’s evolution, we are also still stuck on the question of how intelligence emerges in biology. How can a biological system ever generate consistent, goal-oriented bottom-up behavior when there is no external designer?
In fact, intelligence – a deliberate response to available information, often anticipating the future – is not limited to the minds of certain privileged species. It is distributed throughout biology, at many different spatial and temporal scales. It’s not just smart people, mammals, birds, and cephalopods. Intelligent and determined problem-solving behavior can be found in parts of all living things: individual cells and tissues, individual neurons and neural networks, viruses, ribosomes and RNA fragments, up to and including ‘motor proteins and molecular networks. It can be argued that understanding the origin of intelligence is the central problem in biology, a problem still wide open. In this article, we argue that advances in developmental biology and neuroscience now offer a promising avenue for showing how the architecture of modular systems underlies evolutionary and organic intelligence.
Biologists are trained to focus on the mechanisms of living systems and not on their purpose. As biologists, we are supposed to determine the “how” rather than the “why”, pursuing causality rather than goals. The “why” is not only always present, but it is precisely what motivates the choice of specific “how”, allowing organisms to survive by selecting and exploiting specific mechanisms in an astronomically vast space of possibilities. In the case of the human eye, for example, the optical properties of the lens are only meaningful if they help focus light on the retina. If you don’t ask why the lens is transparent, you will never understand its function no matter how long you study how it becomes transparent.
Indeed, the problem of understanding how intelligence emerges worsens with the “omics” revolution, which generates systematic and quantitative data on genomes, transcriptomes, proteomes and connectomes. Biological systems are dissected into their ultimate complexity, but no magical answer appears at the end of the tunnel. The race for big data does not provide a better explanation of living systems. If anything, that makes it more difficult.
Modern biology faces a lack of fundamental knowledge when trying to explain meaningful and intelligent behavior. How can a system made up of cells and electrical signals generate a well-adjusted body with behaviors and mental states? If cells are not intelligent, how can intelligent behavior emerge from a distributed system made up of them? This fundamental mystery permeates biology. All biological phenomena are, in a sense, “group decisions” because organisms are made up of individual parts – organs, tissues, cells, organelles, molecules. What properties of living systems allow components to work together towards higher level goals?
A common solution emerges in two different fields: developmental biology and neuroscience. The argument takes place in three stages. The first is based on one of the first and best design ideas in natural selection: modularity. Modules are autonomous functional units like apartments in a building. Modules implement local goals which are to some extent self-sustaining and self-monitoring. Modules have basic problem-solving intelligence, and their relative independence from the rest of the system allows them to achieve their goals despite changing conditions. In our example of a building, a family living in an apartment could continue with their normal life and pursue their goals, for example by sending the children to school, regardless of what happens in the other apartments. In the body, for example, organs such as the liver function with a specific low-level function, such as controlling nutrients in the blood, in relative independence from what is going on, say, in the brain.
The second step of the argument is that the modules can be assembled into a hierarchy: Lower level modules combine to form increasingly sophisticated upper level modules, which then become new building blocks for even higher level modules, and so on. In our building, families could belong to a local association, such as a local branch of a political party, whose goals could be to ensure the future well-being of all families in the area. And that party could belong to a parliament, the purpose of which could be to shape the politics of the whole country, and so on. In biology, different organs might belong to the same body of an organism, the purpose of which would be to preserve and reproduce, and different organisms might belong to a community, such as a beehive, whose purpose would be to maintain a stable environment for his members. . Likewise, the metabolic and local signaling goals of cells integrate towards a morphogenetic outcome of complex organ construction and repair. Thus, an increasingly sophisticated intelligence emerges from the hierarchies of modules.
This may seem to solve the problem, except that hierarchical modularity still doesn’t explain how evolution, only changing one item at a time at a lower level, can ever manipulate higher levels. Since upper levels are built with lower levers, wouldn’t you still need to tweak a bunch of things at the same time to change a higher level mod? A third step in our argument addresses this problem: each mod has a few key elements that serve as command buttons or trigger points that activate the mod. This is known as model completion, where activating one part of the system turns on the whole system. In our apartment building, the family would have a central figure, say one of the parents, who would represent the family at meetings and involve them when needed. These trigger points serve to represent the whole module and thus allow these modules to be activated, modified, deactivated or deployed in new circumstances without having to manipulate or recreate all of their parts. Additionally, pattern completion naturally emerges from interconnected element systems with interactions between elements.
In recent years, researchers have found evidence for the completion of patterns in neural circuitry and developmental biology. For example, when Luis Carrillo-Reid and his colleagues at Columbia University studied how mice respond to visual stimuli, they found that by activating as few as two neurons in the middle of a mouse brain, which contains more 100 million neurons,could artificially trigger visual perceptions that led to particular behaviors. These fascinating pattern completion neurons activated small modules of cells that encode visual perceptions, which were interpreted by mice as real objects. Likewise, in work published in 2018, Michael Levin of Tufts University and Christopher Martyniuk of the University of Florida examined data showing how triggering a simple bioelectric pattern in non-neural tissue causes cells to build an eye or other complex organs in new locations, such as a tadpole’s gut.
The idea of hierarchical modularity to explain biological intelligence has already been explored by economist Herbert Simon, neuroscientist Valentino Braitenberg, computer scientist Marvin Minsky, evolutionary biologists Leo Buss, Richard Dawkins and David Haig, and philosopher Daniel C. Dennett, among others. These recent experiments in developmental biology and neuroscience can now provide a common mechanism for how this might work via key nodes that generate model completion. While there is still a lot to learn about how model completion units work, they could provide a solution to the problem of reusing a mod system without having to change everything. The manipulation of local goal-chasing modules, to make them cooperate at multiple organizational scales in the body, is a powerful driving force. It enables evolution to harness the collective intelligence of cellular networks, using and recombining tricks discovered at the lower level while operating robustly despite noise and uncertainty.
Like a ratchet, evolution can thus effectively climb the ladder of intelligence, from simple molecules to cognition. Hierarchical modularity and model completion can help understand the decision-making of cells and neurons during morphogenesis and brain processes, generating well-adaptive animals and behavior. Studying how collective intelligence is emerging in biology can not only help us better understand the processes and products of evolution and design, but could also be relevant for the design of artificial intelligence systems and, more generally, for engineering and even the social sciences.
Printing and recall of cortical sets. Luis Carrillo-Reid et al. in Science, Flight. 353, pages 691-694; August 12, 2016.
The bioelectrical code: an ancient means of computation for the dynamic control of growth and form. Michael Levin and Christopher J. Martyniuk in BioSystems, Flight. 164, pages 76-93; february 2018.
Behavior control visually guided by holographic recall of cortical sets. Luis Carrillo-Reid et al. in Cell, Flight. 178, no. 2, pages 447-457; July 11, 2019.
The computational limit of a “self”: developmental bioelectricity stimulates multicellularity and scaleless cognition. Michael Levin in Frontiers in psychology, Flight. 10, article n ° 2688 December 2019.