What is Life? How does it work?
Is it possible for us to truly understand what life really is -- and how it works, in general, not only for a few special examples? And to understand self-organization and emergence (which include life and free energy as special cases) in a more general mathematical way?
Back at the time of Charles Darwin, before his revolutionary new ideas became popular, biologists often seemed to be doomed to a life of meaningless leaf-collecting. They seemed to be collecting a gigantic archive of raw, descriptive data – without being able to generalize or to predict or to understand the underlying patterns. This has changed a lot since then -- but even today we face a big challenge in trying to generalize and understand all the data which has been accumulated.
Is it even possible to develop a kind of unified, mathematical understanding of life in general? Is it even possible to develop the same kind of universal, coherent mathematical understanding here as it is in the case of physics or in the case of intelligent systems? Is it just plain inherently intractable – or are there ways to deepen our understanding that we simply haven’t yet discovered? Can we develop an understanding general enough, for example, to actually predict what kinds of life might exist in other chemistries or other hypothetical universes? Or even what kind of life we might accidentally create on earth? Can we even be sure, theoretically, about what kinds of scenarios from science about alternative life forms would actually be possible, in some environments? (I have wondered at times about Philip Jose Farmer’s scenario in To Your Scattered Bodies Go.)
such an understanding is impossible – but before we give up, we should
think about how impossible it once seemed to develop a unifying mathematical
understanding of basic physics or of intelligence. The first great unifying
mathematical understanding in physics came from
Life, in the most general context, is really just one part of the larger subject of self-organization or self-organizing systems. Back in 1994, Karl Pribram invited Ilya Prigogine, myself and a couple of dozen others to address the basic questions about self-organization in one of his workshops. He invited me to write a general introduction to self-organization, which was printed next to Prigogine’s paper in the start of the book:
P.Werbos, Self-organization: Re-examining the basics and an alternative to the Big Bang, in K.Pribram, ed, Origins: Brain and Self-Organization, Erlbaum (1994). (“Big Bang?” you ask…)
2006: Click here to see the slides from a more recent talk for mathematicians, where I summarized new work showing how a remarkable hierarchy of self-organization can emerge even from simple, traditional “classical field theories”. In fact, I argue with equations that the incredible richness of everything we see in nature, from elementary particles to life and quantum physics and powerful forms of mind, could all be understood as the emergent outcome of such a theory. I started the talk by saying – no, I am not an extreme conservative, lobbying for a simple form of objectivism. Rather, I believe that science demands a kind of careful step-by-step approach, where we fully exhaust the power of the simplest kind of theory we can construct, consistent with all the data, before we add complexity; we should add complexity only as we understand exactly what kind of complexity we really need. Also, the more I understand how rich the emergent behavior can be even from a simple system, I do take very seriously the possibility that it might just turn out to be the whole story. Who knows?
I have not had time to really focus on this area since then, but I have time to muse a bit on some of the key questions. I have wondered: can we somehow unify our understanding of life with whatever can be learned from general-purpose thermodynamics, nonlinear system dynamics, and artificial life and so on? Could the mathematics developed for intelligent systems be of some use here, if only to suggest more orderly chains of approximation or ways to handle genomic/proteomic data? Von Neumann certainly included this area as one of the three big areas he focused on, in trying to develop new mathematics to help us better understand the foundations of everything.
NSF for many years ran a small activity called “Quantitative Systems Biology,” which comes the closest I know of to providing a venue for answering these kinds of questions. The Engineering perspective is extremely important here, as in the study of intelligent systems, because it stimulates us to think about the possibilities for life in general, not just the particular list of life forms we happen to have access to right now.
Greg Bear’s books about Darwin’s children – fictional as it is in many ways – acutely poses a related question here: to what extent is the “junk DNA,” 97% of the genome, actually a kind of learning system, like the brain. More precisely, how much of the genome is intended to help us learn how to choose better gene expressions, as opposed to merely specifying the final “actions” at the output layer of the “gene brain”? Certainly “metalearning” is a crucial unmet challenge to evolutionary computing, as well as to more brain-like approaches to nonconvex function maximization. Metalearning is only one of several theories about the function of the “junk DNA” being discussed in the the new International Postgenetics Society. It seems very odd to me that even today there are many who truly believe that “junk DNA” has no function at all – that 97% of the genome is all just a matter of gross inefficiency and accidental waste!
Would it be possible to use approaches more like quantum statistical thermodynamics
(like the master equations so familiar in quantum optics) or even simple eigenvector and
growth analysis, to try to solve for the forms of life that emerge as simple or unicellular
patterns in a homogeneous fluid environment? Would the early Russian work on autocatalytic reactions help point the way here? (Of course, autocatalysis, aging and
death are likely to be important even to the most general initial models…)
A few of my more recent thoughts in this area have appeared in:
W. Freeman, R. Kozma and P. Werbos, Biocomplexity: adaptive behavior in complex stochastic dynamical systems, Biosystems 59, p.109-123, 2001.
P. Werbos, From the termite mound to the stars: Meditations on discussions with Ilya Prigogine, Problems of Nonlinear Analysis in Engineering Systems, No.3(19), vol. 9, 2003.
P. Werbos, Extending Chaos and Complexity Theory to Address Life, Brain and Quantum Foundations, In IJCNN 2000 Proceedings, IEEE, 2000.
Links between sociobiology and my own view of intelligence are discussed in:
P.Werbos, Values, Goals and
Utility in an Engineering-Based Theory of Mammalian Intelligence, in
Karl H.Pribram, ed. , Brain and Values, Erlbaum:
In the field of history, many theorists have argued that a “frontier” – a high rate of growth, as one finds in a kind of open system – is essential to maintaining high civilization, altruism, freedom, progress, etc. Yet an interesting aspect of sociobiology is that this parameter – the degree of growth of the entire population – is not really such an important driver as often thought. George Gaylord Simpson has described related, but more complete, versions of the frontier concept. But in the end… there are many parameters which come into it. The stable historic pastoral societies which E.O. Wilson describes sounds like a crucial ingredient to some kind of progress… until one imagines turning them into tribes in Afghanistan carrying nuclear weapons and growing opium.
Clearly there are many connections here yet to be fully understood…