What is a TALKING OCTOPUS?
Our predominant theories assume that the organized appearance of life and the reliable recurrence of ontogeny is due to genomic influences. In this view, the genome even carries responsibility for producing learning. Learning is thought to be present only in complex structures like brains, and it is assumed to be aimed at solving problems related to survival and reproduction. I believe this is wrong on all three counts. The theory of life, including Darwinian natural selection, only makes sense if a form of learning is present in every living cell that aims to find new problems and solve them. This form of learning can’t be generated by genomics because it purposefully changes the deployment of heritable material, using it for its own contingent, adaptive ends, and inheritance would be impossible without it. I’ve named this unknown mechanism “epistolution,” to distinguish it from other concepts of learning. Although all ontogeny requires epistolution, especially clear examples include embryonic development, wound healing, regeneration, cancer, memory, dreaming, creativity, swarm intelligence, epigenetic inheritance, and the placebo effect. These essays are an attempt to clarify epistolution so that it can be studied in detail in laboratories, and to imagine what exactly the societal consequences will be when we decipher this unknown mechanism. The “talking octopus” will be the first really impressive implementation of this principle; unlike todays algorithmic mimicry of intelligence, they will be the first creative, intelligent, moral machines on earth.
Important Points
1. Learning is viewed as an inherited trait, but gene expression is controlled by an interaction between the phenotype and the environment. This makes genomic function a consequence of learning as well as a cause of it. Learning cannot be simply an inherited trait under genomic control.
2. It requires at least some learning to get from any genotype to any phenotype. Plasticity and adaptation to changing conditions is a basic feature of all embryonic development, even the development and reproduction of single cells. For example, take E. coli clones and put them in different environments, and you will see that they always adapt in modest ways to accommodate different conditions, and that their adaptation logically has consequences for gene expression.
3. This modest physiological adaptation is a feature of all living systems no matter how simple. If so, not only is this sort of learning not an inherited character, it may be a necessary prerequisite for ontogeny and therefore inheritance. If it is present in all life, it would appear that natural selection might not be possible without it. This would mean some form of learning was already present in the first life form.
4. It is tautological that without survival and reproduction a lineage could not persist, but this doesn’t prove that life is required to exclusively seek out survival and reproduction, it could also happen as a byproduct of pursuing another motive. A universal form of learning presents a motive, an organizing principle for life, that could theoretically supercede survival and reproduction.
5. Learning is currently defined and tested in terms of problem-solving ability, but this was based on the assumption that learning was an inherited trait which arose to solve problems related to survival and reproduction. We have established above that this may not be the case. Such tests exist not in the cognitive domain of the organism but rather in the domain of the observer. If learning is not an inherited trait, it cannot be targeted primarily at solving survival or reproductive concerns because it hasn’t been naturally selected. Organisms have experiences, gain knowledge, and accumulate memories, but these cannot possibly be “about” survival and reproduction because they have never experienced nonsurvival or alternative rates of reproduction. Instead, all organisms may be trying to actually understand the world themselves, which means finding their own problems to solve.
6. Unlike problem-solving, which presupposes a goal implanted by natural selection, understanding means open-ended experimenting to develop a sense of what entities and causal forces exist in one’s surroundings. This would mean that all life, every living cell and also every whole organism, would contain some representation of the world that conflicts, in some respect, with experience. Understanding could thus serve as an intrinsic motivation for behavior.
7. If such understanding is universal then all cells have, if not their own consciousness, at least their own frames of reference, memories, and aims. These mental attributes could determine what and how organisms learn, develop, and what they seek to accomplish. Physiological development and behavior could be guided primarily not by the need to solve problems related to survival and reproduction, but by the need to find problems that help an organism gain understanding.
8. There must be a principle, though it is still unknown, that keeps life more organized than nonlife. Our predominant theories assume that this organized appearance and the reliable recurrence of ontogeny is due to heritable (read genetic or cytoplasmic) influences. In its extreme form, this view sees control exerted by a plan or code received from evolution by natural selection as a form of transcendent intelligence. Since what is inherited is a matter of contingent historical accident compounded over trillions of generations, according to this account there is no reason to believe that any general principle could be extracted that might elucidate how learning (and thereby ontogeny) occurs. Life would be intractably complex. As Branscomb writes the “system and its historical antecedents (would be) mechanistically and causally inseparable.”
If, however, we conclude that no inheritance would have been possible without some learning, then there must be a general endogenous mechanism according to which all living beings learn because it has been shared with all the ancestors of the first living cell, and it is what has allowed inheritance to proceed.
9. This principle stands in sharp contrast with the assumption that the aim of life is autocatalysis or autopoiesis. The chemical self-maintaining capacity of cells can be explained by their historical inheritance and by natural selection; after all, it is tautological that what survived is still present. As we have shown above this isn’t sufficient to explain ontogeny. Epistolution presents a new principle that would give life organizational closure in a different way by establishing a boundary within which subjective causal representations are formed and informational input acquires semantic meaning.
10. So what is the mechanism, common to all cell types on earth, that might produce this universal learning mechanism that activates a genotype in order to produce a phenotype? Could it be oscillators and servomechanisms? In cybernetics, life was considered as a set of servos and oscillators driven by the genome, but in this account, it is epistolution that drives them. In light of this new insight, a universal principle might be derived that would configure a set of oscillators as a learning mechanism interpreting novel data rather than as a readout of evolved, inherited signals. This would resolve the problem of why and how genotypes lead to phenotypes.
A formalized model of epistolution might serve as a general principle explaining life. If so, this conceptual breakthrough would lead to more effective medical intervention and regenerative therapies, and eventually to the ability to design and build organisms fit for “artificial” purposes. Epistevolving machines might be built that, in contrast to the current statistical mimicry of intelligence, could gain causal understanding of the world by inventing and testing subjective theories. This might comprise a powerful new technology amplifying scientific, technological, aesthetic, and moral progress in all intellectual domains.