On Thursdays at noon Yhouse holds a lunch meeting at the Institute of Advanced Study, in Princeton. The format is a 15 minute informal talk by a speaker followed by a longer open-ended discussion among the participants, triggered by, but not necessarily confined to, the topic of the talk. In order to share I am posting a synopsis of the weekly meetings.
Synopsis of Randall Beer’s YHouse Luncheon talk 10/19/17
Presenter: Randall D. Beer (Indiana University)
Title: Organisms As Integrated Wholes
Present: Randall Beer, Piet Hut, Yuko Ishihara, Kim Chung, Steven Lim, Ohad Nachtomy, Jeff Ames, Harvey Lederman, Olaf Witkowski, Eran Agmon, Michael Solomon
Abstract: "I will briefly describe two intertwined research programs. The first concerns issues of embodiment, situatedness and dynamics in understanding how an animal's behavior arises from the interaction between its nervous system, its body and its environment. Specifically, we use genetic algorithms to evolve model brain-body-environment systems and then analyze their operation using the tools of dynamical systems theory and information theory. This approach has been applied to a wide variety of behaviors, including locomotion, action-switching, learning, categorization, selective attention, and referential communication. The second concerns the organization of minimal living systems and its consequences. Specifically, we analyze persistent spatiotemporal entities in cellular automata models from the perspective of autopoiesis and enaction. We identify the local processes that underlie these entities and show that they form self-producing and self-individuating networks. We then divide the environmental perturbations that such an entity can receive into destructive and nondestructive subclasses and characterize the structure of the entity's responses to all nondestructive perturbations. Finally, we derive the contingent structure of sequences of nondestructive perturbations, effectively mapping out all possible "lives" that such entities can live."
Dr. Beer began by describing his two main research interests. The first is mechanisms of animal behavior. The second is understanding the organization of living systems. In the first, he looks for the mechanisms of behavior in the closed-loop interaction between brain, body and environment. He uses evolutionary algorithms to generate model brain-body-environment systems and the mathematical tools from dynamical systems theory and information theory to analyze them. He has looked at a variety of behaviors, ranging from locomotion to minimally cognitive behaviors such as selective attention, learning, and communication. Some of these models are empirically grounded and some are more abstract. An example of the former is work on modeling C. elegans is a 1 mm long soil living round worm. It is unique in that the connectivity of its nervous system has been fully mapped. Because it is extremely difficult to do electrophysiology directly, he and his collaborators use genetic algorithms to find possible values of the neural parameters that produce behavior consistent with the animal.
His second research interest is to understand the organization of living systems. What makes them biological systems in the first place? Some answer biochemistry and look at specific molecular reactions. This has been used in the search for extra-terrestrial life. Some answer evolutionary properties. Instead, he looks at Autopoiesis and Enaction – the organization of processes that distinguish living from non-living systems. He uses models in which the universe is divided into states following universal laws, and he then analyzes how higher order structure arises that instantiate an entire system. Once you have self-producing, self-perpetuating autonomous processes, then you can look at environmental perturbations and the effects of the perturbations. Can the entity survive? How does the entity evolve?
He ended by saying that although these two programs sound different, they are related. Both look at living systems as integrated wholes. How does an integrated system come into being? Both share a methodological approach: construction of essential models that are stripped to only essential characteristics. As in physics, once you have a toy model, you take it seriously. That is, you may not believe the model is true, but you try to extract all the information you can from that model.
The Questions and discussion began with Piet recommending Randy’s paper on the organization of life as engineering philosophy. THE REFERENCE FOR THIS PAPER:
Beer, R.D. (2014). The cognitive domain of a glider in the Game of Life. Artificial Life 20(2):183-206.
Beer, R.D. (2015). Characterizing autopoiesis in the Game of Life. Artificial Life 21(1):1-19.
Q: Piet asked, “How much response to your work have you gotten from the community?”
A: RANDY: He has two papers of note written in backwards order. 1) His paper describing “If you have a localized bounded self-perpetuating entity then you can map out all possible lives it could live in all possible universes”, got a very good response. 2) His “How the entity arises from the underlying physics” paper had less impact at least for now. There is disagreement in the research community about whether the model fits. Some argue you should start with the thermodynamics.
Q: Piet: Has everyone here heard of Autopoiesis?
Q: Ohad: Is the entity individuated by the system or environment?
A: RANDY: Autopoiesis literally from the Greek means “Self Creating”. But, the term also includes the notion of “Self-Individuating”. The concept originated with two Chilean biologists Humberto Maturana and his student Francisco Varela. They defined the cell as the unit of life. But there are too many definitions of Life and no definition is sufficient as they all rule out aspects such as biodiversity. The two aspects of the living cell are 1) Networks of chemical properties that generate the components of the network itself. 2) The Self-individuation part is the network specifies its own criteria of distinction. An example is cell membranes that are composed of lipids that the cell produces and that maintain the cell integrity, separating inside from outside. In his work, Randy must identify a network of processes.
Eran: Proteins do not occur naturally, but they create proteins which make proteins. Cells can take in compounds and produce the components it needs.
A: Randy: Replication, which is often cited as a requirement for life, is not an essential property but is derived secondarily inside of Autopoiesis.
Q: Piet: In math, physics, chemistry you can define what you work with. In biology, you cannot. There is no good definition that distinguishes alive from proto-life or non-living.
A: Randy: You take a formal, toy, concrete model and then study it according to some definition of life. Common definitions include biochemical, reproductive, other properties that may not apply. Is a sterile animal alive? Of course.
Q: Michael: With regard to a property common to all living things, and with regard to membranes, I have read that the Gibbs Free Energy across all biologic membranes is a constant, and must be between minus 50 and minus 60 kJ/mmole. Is that a constant characteristic of life on earth, and why should that be?
Eran: That transmembrane potential varies from one cell to another. It is only in mitochondria, which require a constant electrical gradient across the membrane to function, that the free energy is fixed.
Michael: I am not referring to transmembrane electrical potentials, but to Gibbs Free Energy, the potential ability to extract work.
Randy: That would include at least both electrical gradients and osmotic concentration gradients.
Eran: I was thinking only of electrical gradients, not of Gibbs Free Energy. Life as we know it exists in a very specific thermodynamic system. That does not apply in the Game Of Life.
Randy: The Game Of Life is the cellular automata model that he uses to model life as he has described.
Q: Yuko: Can you clarify Enacting?
A: Randy: Historically, Maturana and Varela focused on mechanisms in physical systems. But later, Varela subsequently brought in phenomenology into the model and in the title of his book he called this Enaction.
Q: Ohad: Is the basic living unit something that can only exist in relation to the environment?
A: Randy: We are referring to Functional not as structure but as processes.
Q: Ohad: That is very much as Kant says.
Eran: Varela’s paper from 2000 is titled “Life after Kant”
Ohad: Kant says every organ must contribute to the whole organism and vice versa.
Eran: I have the quote from The Critique of Pure Reason here on my laptop: “In such a product of nature every part not only exists by means of the other parts, but is thought as existing for the sake of the others and the whole, that is as an organic instrument.”
A: Randy: There are other contemporaneous theories similar to autopoiesis, including Tibor Gánti’s “Chemoton” (discussed in his 1971 book as the original ancestor of all organisms. MJS) or the ideas of Robert Rosen. The book that first introduced Randy to autopoiesis was “Understanding Computers and Cognition” by Terry Winograd (published in 1986).
Q: Michael: Many of the process of living things seem circular, such as DNA requiring proteins that are coded in the DNA. Is there any way of determining the sequence in which these processes arose? Chicken or Egg?
A: That’s historical. He is working on not just our particular life form, but on how the parts in any life form assemble spontaneously.
Eran: Which came first is the wrong question. Ask instead, how did the parts assemble themselves?
A: Randy: In the world, there are contingencies. In the model, he can start with different contingencies and can vary the contingencies. Further, a modern cell may look nothing like the first living system.
Q: Ohad: Must there be an individual in origin of life?
A: Randy: That depends on your definition of life.
Q: Olaf: You could have models without an agent.
A: Randy: You would need to specify a definition in order to formalize it.
Q: Piet: Life as we know it is metabolic – something eats something else. But there could be life based on cognition. The first cell had to choose what to eat and what not to. You could imagine the sci fi story of Solaris in which the entire ocean is alive. The cell is the first time the universe was split into subject and object – inside and outside the cell.
Eran: Cells are very persistent. If a cell is trying to divide and one pathway is blocked, the cell will use a different pathway to continue to divide. He uses the word “try” although that may not be the correct word.
A: Randy: For Maturana every living cell is also a cognitive agent.
Q: Lim Chung: On integrated whole and units, aren’t we speaking of levels on which we can zoom in or out. What levels are we discussing as living units?
A: Randy: Whether multicellular organisms are capable of autopoiesis is controversial. Varela calls multicellular systems “organizationally closed”.
Eran: Is a corporation a living system? Members don’t create one another.
Piet: Of some 50 plus characteristics we may consider in defining life, he likes to use, “Building reliable systems out of unreliable components”. Unlike a building where the bricks can be reused if the building collapses, in life the whole is more reliable than its component parts. My cells may die but I persist. This is also true of computers.
Q: Lim Chung: Is there a line we can draw to define levels? The closure idea seems artificial.
A: Randy: Closure is really only in mathematical models. In real systems there are other contingencies.
Piet: Biology studies only a single instantiation of a thing, Life. When we find other forms of life we will have more to choose from.
Eran: Even on earth there are some unusual life forms. There is a tree that has an enormous root system extending for miles and individual trees grow from that root system and may die and others grow, but the tree is a single organism. In our own bodies, 90% of the DNA is not human DNA but belongs to the microbiome of bacteria that live in our intestine, skin, and elsewhere and that are essential for our survival.
Piet: All life that we know is derived from a single community of ancestors. But in the universe, there may be any number of other forms that we have yet to discover.
Olaf: Perhaps Information does not need to be embodied. We could look at pure information. Consider the Biosphere as transmitting information.
A: Randy: You don’t have subject/object without an individual. Then you are just looking at the universe and not at a system from which higher order structure emerges.
Eran: Subject/Object should not be determined by the subject.
Olaf: Take a network with boundaries, and then another network that links to the first. Then take N number of networks interfacing. This model can evolve and result in a reality (a perspective) that may have no relation to actual physical reality.
A: Randy: This can be modeled in detail in the Game of Life.
Piet: If you imagine the planet where the entire surface of the planet is alive, then that becomes the entity.
Olaf: Subject/objects can be temporary.
Piet: This subject/object split discussion might be a paper.
We continued the discussion later considering Olaf’s proposal that we could play a game in which we each transmitted a set of instructions to another player, and then further instructions could be shared. We are transmitting information but the “agents” of that information are not defined. Could that serve as a model for a form of life without an entity? Eran suggested the model of a navy ship in which the pilot and captain send orders around the ship to sailors who carry out the orders and report back or send information to other sailors who carry out their own activities. In fact, there is no agent involved in that model either. (So, who is responsible when the ship collides with another?) It was not clear how this model would apply to formulate a living system that has no individual or agent, but only pure information.
Michael J. Solomon, MD