The Definition of Life

The aim and the end of all becoming is the development of potentiality to actuality, the incorporation of form in matter.
—Aristotle, attributed by Jeremy Campbell, in Grammatical Man

And it is in no way possible for anything to be responsible for its own generation and decay. For the mover must preexist the moved, and the begetter the begotten. But nothing is prior to itself.
—Aristotle, De Motu Animalium, trans. Martha Craven Nussbaum

The cause of the origin of a thing, and its eventual utility, its actual employment and place in a system of purposes, lie worlds apart.
—Friedrich Nietzsche, quoted in Daniel C. Dennett, Darwin’s Dangerous Idea

Contents

• Defining Life
• Criteria for Our Definition
• Simple Replication
• Von Neumann Replication
• Resistance to Entropy
• Autopoiesis
• Responsiveness
• Continuance Through Others
• Homeostasis
• Patterns of Complexity
  • Algorithmic Complexity
  • Logical Depth
  • Effective Complexity
  • Ranges of Fluctuation
• Behavior
• A Matter of Degree

Defining Life

Since life is such a ubiquitous and fundamental concept, the definitions of it are legion.
—John D. Barrow and Frank J. Tipler, The Anthropic Cosmological Principle

Words are our servants, not our masters. For different purposes we find it convenient to use words in different senses.
—Richard Dawkins, The Blind Watchmaker

What is life? And why should we care? Well to begin with, we are living beings, and that fact distinguishes us from most things in the Universe. Though humans are not the only living things, we are among the few, so understanding the nature of life might be an important step toward understanding ourselves.

As Richard Dawkins points out, people define life in different ways for different purposes. For everyday situations, it seems to me that we have a common-sense set of criteria, somewhat along the lines of:

Does it look like a person or an animal? If so, is it moving? Does it respond to being spoken to or touched? Failing this, is it breathing, or is its heart beating?

Does it look like a plant? If so, does it have green leaves? If not, could it be because it is winter? . . . Etc., etc.

The question is, can anything meaningfully be done to define life that would not simply be a repetition of such everyday notions. Once upon a time, philosophers like Plato believed that things in the everyday world are imperfect reflections of perfect forms or concepts in some type of higher realm. In defining a class of objects, we would be searching for an understanding of that mystical essence that inheres in all of them.

But as David Hume said, somewhat more recently,

Nothing is more usual than for philosophers to encroach on the province of grammarians, and to engage in disputes of words, while they imagine they are handling controversies of the deepest importance and concern.

Even more recently, the logical positivists have stressed the distinction between analytic and synthetic propositions. An analytic proposition is one that is true by definition, such as saying that "men are adult male humans." Such a statement is an assertion about words. On the other hand, a synthetic proposition asserts some new truth about the world that was not inherent in the words themselves. Synthetic propositions are subject to verifiability; we can perform experiments or other observations to determine whether these propositions are true.

A third class of propositions consists of statements that are neither analytic or synthetic, and to the logical positivist, such statements are simply without sense. The logical positivists sought to expose much of traditional philosophy as meaningless discourse, a sort of neurotic disease.

Which of these three types of statements—analytic, synthetic, or meaningless—am I planning to make about life? Well, neither of the first two. Whether my statements will fall into the third category remains to be seen.

I would argue that it would be a good thing to have a workable abstract definition of life, and that such a definition need not be wholly arbitrary, but can be defensible to a degree. However, the point is not to describe some sort of metaphysical essence of life. Rather, the point is to define life so that the term can be usefully extended to situations we have never before encountered.

For example, the legal definition of "death" now has to take account of situations that have only recently been made possible by medical science. Nowadays many comatose people can be kept on life support machines for years. But are we to regard this as life, or specifically as human life, with all the ethical considerations that we attach to human life?

But the issues reach beyond ethics and into the area of scientific discourse. Researchers such as James Lovelock have recently developed the Gaia theory, which is the notion that the earth as a whole is a living organism. The idea is vigorously disputed by many other scientists. But the real question is, how scientists can even have a debate about whether the earth is alive, if they haven’t agreed in advance what they mean by "alive"? Lovelock, to his credit, does make an effort to define his concept of life, but states his conclusions in only the most incomplete and tentative sort of way.

The idea of Gaia is said to have been sown in popular culture by the photographs of green-and-white Earth taken from the barren moon. Similar changes of perspective may result by our explorations of other planets, such as Mars. If we found life there, could we recognize it? After first noticing it, could we agree on whether it was indeed life or not?

The possibility of meeting extraterrestrial life may seem a distant one. Lovelock himself has provided good reasons for believing that there is no life currently on Mars, whether or not there was in the past. And the density with which life is scattered through the galaxy or the universe as a whole is almost totally unknown. We might make contact tomorrow or never. (Sometime in between would be my guess!)

But there are two much more immediate trends on our own planet that are forcing us to stretch and readjust our definition of life. These are trends in biotechnology and information technology.

The Human Genome Project is currently mapping the locations and function of all the genes in human beings. It is difficult to fully appreciate the significance of this project. With such information in hand, or even a small part of it, we will attain the ability to redefine ourselves in ways that we have not yet begun to imagine. Clearly, we will need to generalize our concepts of what life is unless we are to suppose that future generations, who may differ from all life today in major ways, are not "alive."

Meanwhile, cognitive scientists are continuing a three-decade long effort to use computers to model and possibly someday to embody intelligence. So far, their efforts have mainly taught us that the task is more difficult than anyone had imagined. But arguments that artificial intelligence is impossible tend to lapse into mysticism of the worst sort. The great conceptual gap here arises from the fact that most of our own information processing is below the conscious level. From this fact arises the deeply-felt intuition that thought comes from an inherently magical and nonmaterial place. Even our conscious experience bears no resemblance to the cellular and molecular processes that apparently give rise to it. But it doesn’t follow that our experience is not made of cells and molecules. It simply means that there is a big difference between being a process and looking at the process from the outside.

I don’t have a timeline for the creation of artificial intelligence, but look at it this way. It took Nature something like three billion years of trial and error to generate human beings. Shouldn’t human beings, with the benefit of planning and foresight, be able to duplicate and exceed this feat of blind Nature in a mere fraction of that time span?

If we succeed in creating intelligence, will it follow that we have created life? Is intelligence even possible without life? I will argue that it is not (but only after first placing the definition of life an abstract level that does not inherently exclude computer chips).

One thing almost everyone would agree on is that life is a complex phenomenon, with many facets that emerge only after careful examination. I once bought a book on identifying trees, and was surprised to discover that I had never really looked at a tree before I had that book. What I mean is, I had never looked at a tree before at that level of detail, or with that much appreciation for how trees vary. Now, you could argue that that book was simply like a dictionary, a list of agreements about how we should name plants. But I know that it was more than that. For myself, and for the people who wrote it, it was a voyage of discovery, a dawning awareness of wonders that would otherwise have gone unnoticed.

Let us therefore attempt to identify life. Let us make a guidebook that we could take on a field trip to any planet, either elsewhere in the galaxy, or later in our own history, and use to direct our observations. Only when we know what to look for can we truly see. And, though we take this voyage only in imagination, we are not likely to return unchanged.

Criteria for Our Definition

Rather than add to the already unmanageable list of definitions, we shall simply give what seem to us to be the sufficient conditions which a lump of matter must satisfy in order to be called "living." We shall abstract these sufficient conditions from the various definitions proposed over the last thirty years by biologists.
—John D. Barrow and Frank J. Tipler, The Anthropic Cosmological Principle

Whenever biologists try to formulate definitions of life, they are troubled by the following: a virus; a growing crystal; Penrose’s tiles; a mule; a dead body of something that was indisputably alive; an extraterrestrial creature whose biochemistry is not based on carbon; an intelligent computer or robot.
—William Poundstone, The Recursive Universe

We will be searching for a definition of life that is useful. In order to be useful, the definition should meet the following criteria, so far as possible:

• Sufficiency. It should provide the sufficient conditions that enable us to specify whether something is living or not.
• Common Usage. These conditions, when applied to "easy" examples, should classify those examples in the same way we normally do. Easy examples include obviously living things such as people, animals, plants, and bacteria; things that were alive but are now dead; and things that we would never normally consider alive, such as rocks, screwdrivers, and growing crystals.
• Extensibility. It should be possible to apply these conditions to "difficult" examples with some kind of coherent result. Difficult examples include viruses, mules, fire, simple feedback systems (such as those with thermostats), Gaia, extraterrestrial creatures, and robots.
• Simplicity. The definition should be as simple as possible, with a minimum of ifs, ands, or buts.
• Objectivity. The definition should refer to measurable and objective properties of the organism. That is, the definition should be specific enough so that different people can be counted on to apply the definition in the same way when they encounter a new "difficult" example.

Once we have determined the sufficient conditions that something must satisfy in order to be considered living, we can go on to ask what additional properties follow from the fact that something is alive. Life may involve certain engineering problems that have only a finite number of possible solutions. Other engineering problems may have a variety of possible solutions, including many that have not been used by any familiar life forms.

Our procedure will be to review some of the definitions that other authors have given, beginning with the most naive and progressing to the more satisfactory. Then we will try to improve on the best existing definitions. 

Simple Replication

Virtually all authors who have considered life from the point of view of molecular biology have regarded the property of self-reproduction as the most fundamental aspect of a living organism.
—John D. Barrow and Frank J. Tipler, The Anthropic Cosmological Principle

Barrow and Tipler thus begin their discussion of life with a statement that is historically untrue, as we shall see later; not all molecular biologists regard self-reproduction as the most fundamental or defining aspect of life. And by beginning this way, Barrow and Tipler commit themselves to a particular type of definition without stopping to consider whether it is really necessary.

Almost immediately this definition runs into trouble, because on the one hand, there are easy examples of living things that do not or cannot reproduce; and on the other hand, there are easy examples of non-living things that do reproduce.

Barrow and Tipler are aware of these difficulties, and they begin by discussing the living things that do not self-reproduce: among them, childless people and mules. (Mules, of course, are the offspring of horses and donkeys, and cannot have offspring themselves.) Then the authors present the following rationale:

But such creatures are metazoans, which means that they are all composed of many single living cells, and generally each cell is itself capable of self-reproduction. Many human cells, for instance, will reproduce both in the human body and in the laboratory. In general, all known forms of living creatures contain as sub-structure cells which can self-reproduce, or the living creatures are themselves self-reproducing single cells. All organisms with which we are familiar must contain such cells in order to be able to repair damage, and some damage is bound to occur to every living thing . . . The ability to self-repair is absolutely essential to a living body.

Since all living things are largely composed of cells which can self-reproduce, or are autonomous single cells with self-reproductive capacity, we will say that self-reproduction is a necessary property which all living things must have at least in some of their substructure.

What the authors have done at this point is to jump to a different level of analysis. At the start, they were talking about self-reproduction of the organism as a whole, but now they are settling for self-reproduction of the constituent parts. The strangeness of this viewpoint is evident if you consider the following example:

"He La" cells—from the cervix of Henrietta Lane, a woman who lived in Washington, D.C.—continue to be grown in laboratories around the world, despite Lane’s death from cancer of that same cervix in the 1950’s.
—Lynn Margulis and Dorion Sagan, What is Life?

Now, this is an easy example of someone who is not living, but whose constituent parts are continuing to reproduce—continuing for decades, in fact. It is true that HeLa cells are still alive, but Henrietta Lane is not. This fact suggests that one ought not to confuse the properties of an organism as a whole with the properties of the parts that constitute it.

Barrow and Tipler come up with a similarly awkward rationale for dealing with the things that reproduce but are not alive. They mention the examples of salt crystals and mesons, each of which will produce copies of itself under suitable conditions, and state:

Yet we would be unwilling to regard either salt crystals or mesons as living creatures. The key distinction between self-reproducing living cells and self-reproducing crystals and mesons is the fact that the reproductive apparatus of the cell stores information, and the specific information stored is preserved by natural selection. The reproductive "apparatus" of crystals and mesons can in some cases store information, but this information is not preserved by natural selection . . . Ultimately, it is natural selection that corrects errors and holds a self-reproductive process together, as Eigen and Schuster have shown in their investigation of the simplest possible molecular systems exhibiting self-reproduction. Thus, basically we define life to be self-reproduction with error correction.

There are some oddities about the new clause in their definition. First, natural selection doesn’t necessarily perform error correction, although that may be its effect much of the time. Natural selection preserves the organisms best adapted to their environment, not those most similar to their parents. The distinction may not be important in the case of most mutations, which are distinctly negative in their effects (unless their effect is masked by the corresponding gene from the other parent). But if natural selection always worked to perform error correction, evolution would not be possible.

Moreover, there is a more serious objection. Barrow and Tipler have made another jump here to a different type of analysis. Suddenly life is being defined based on a historical criteria. In other words, something is being judged as life not on the basis of what it is or how it behaves, but on the basis of the way it was created. This is a troubling development, because it means that we could not tell whether something is alive unless we also know that it is the product of Darwinian evolution. But people knew that things were alive before Darwin ever proposed his theory!

Additionally, this criteria would exclude life forms created by design. Surely, if we had been created by an intelligent God, it would not follow that we were any less alive?

Following are some of the applications that authors draw from this definition:

Note that a single human being does not satisfy the above sufficient condition to be considered living, but it is made up of cells some of which do satisfy it. A male—female pair would collectively be a system capable of self-reproduction, and so this system would satisfy the sufficient condition. In any biosphere we can imagine, some systems contained therein would satisfy it . . .

A virus satisfies the above sufficient condition, and so we consider it a living organism . . .

Automobiles, for example, must be considered alive since they contain a great deal of information, and they can self-reproduce in the sense that there are human mechanics who can make a copy of the automobile. These mechanics are to automobiles what a living cell’s biochemical machinery is to a virus. The form of automobiles in the environment is preserved by natural selection: there is a fierce struggle for existence going on between various "races" of automobiles! In America, Japanese automobiles are competing with native American automobiles for scarce resources—money paid to the manufacturer—that will result in either more American or more Japanese automobiles being built!

At least the authors are applying their definition consistently. But in doing so, they have shown that it cannot handle the easy examples. For surely a single human being is an easy example of something that is alive, and an automobile is an easy example of something that is not alive. A virus, on the other hand, is a difficult example that is considered a living thing by some biologists but not by others; it is a borderline case at best.

To summarize, the Barrow/Tipler definition of life fails two of the criteria we proposed at the start: it doesn’t handle the easy cases, and it isn’t simple (they have to add special clauses to handle cases like mules and human reproduction).

How did the authors come to settle on a definition with such obvious inadequacies? Were they using different criteria than us for developing their definition? It seems not. Rather, they are aware of the problems with their definition, but they have despaired of finding anything better:

A consequence of giving sufficient conditions rather than necessary conditions is the elimination from consideration as "living" many forms of matter which most people would regard as unquestionably living matter. This situation seems unavoidable in biology. Any attempt to define some of the most important biological concepts results either in a definition with so many caveats that it becomes completely unusable, or else in a definition possessing occasional ambiguities.

Let us proceed through some of the alternatives and see whether we can come up with anything more satisfactory.

Von Neumann Replication

The writer William Poundstone, in a fascinating book called The Recursive Universe, advocates a definition of life based on some work by the great Hungarian/American mathematician John Von Neumann.

Von Neumann was much occupied by the question of how an organism can reproduce itself. But not all types of reproduction were equally interesting. Von Neumann drew a distinction between trivial and nontrivial self-reproduction. We have already encountered some examples of trivial reproduction in the previous section: salt crystals and mesons. Poundstone mentions some further examples, including certain types of tiles invented by geneticist L. S. Penrose. (These are not to be confused with the "Penrose tiles" invented his mathematician/physicist son, Roger Penrose.) If you have a tray full of these tiles, and you shake the tray so they jostle together, they remain a disorganized mess. However, if you link two of the tiles together in a particular way, and shake the tray again, the other tiles start linking together into pairs that mimic that first seed pair. The first "seed" pair can be said to have reproduced.

Remember that Barrow and Tipler added their criterion of "preserved by natural selection" to exclude such trivial examples of self-replication. Von Neumann took a different tack by explicitly defining what he would consider as non-trivial reproduction. To Von Neumann, non-trivial reproduction must involve a "universal constructor," which is a mechanism that can create any number of different things, provided that it has a blueprint to work from.

In the following quote, Poundstone derives the following criteria for living things from Von Neumann’s work:

(1) A living system encapsulates a complete description of itself.

(2) It avoids the paradox seemingly inherent in (1) by not trying to include a description of the description in the description.

(3) Instead, the description serves a dual role. It is a coded description of the rest of the system. At the same time, it is a sort of working model (which need not be decoded) of itself.

(4) Part of the system, a supervisory unit, "knows" about the dual role of the description and makes sure that the description is interpreted both ways during reproduction.

(5) Another part of the system, a universal constructor, can build any of a large class of objects—including the living system itself—provided that it is given the proper directions.

(6) Reproduction occurs when the supervisory system instructs the universal constructor to build a new copy of the system, including a description.

—The Recursive Universe, Chapter 11

Together with Arthur W. Burks, Von Neumann proved that such "self-reproducing automata" are possible. This theory also proved to be a correct, though abstract, model of how real cells reproduce using DNA. So it seems clear that Von Neumann did hit upon something important about the nature of reproduction. The question is whether this model also amounts to a definition of life.

This definition does seem more promising than the Barrow/Tipler definition, in that Von Neumann’s criteria for non-trivial reproduction are functionally based rather than historically based. How does it fare when applied to easy and difficult examples?

The Von Neumann criteria do eliminate some easy examples of non-life, including salt crystals, mesons, and Penrose tiles. These criteria go further than Barrow/Tipler by also eliminating automobiles (because they not include universal constructors). So this definition deals well with some easy examples of non-life.

The definition also gives an interesting answer for one of the difficult examples. Viruses are considered non-living because they also do not include universal constructors; instead, they invade cells and commandeer the universal constructors there in order to reproduce. Thus viruses meet only the first three out of the six Von Neumann criteria. This seems to be an acceptable conclusion for what is usually regarded as a borderline example of life at best.

Unfortunately, the definition does not deal so well with some of the other easy examples. Under this definition, mules would clearly not be living. Poundstone denies this, but by using the same rationale that Barrow/Tipler adopt: he refers to the fact that the mule is made of cells that themselves reproduce. As we have seen before, all that really follows from this observation is that a mule is a dead thing made out of living cells. Thus, I mentioned previously the example of HeLa cells, the living cells of a deceased person.

In fact, in the next paragraph Poundstone asserts that dead bodies in general should be considered living, at least until decay destroys the internal structures of the cells. For, until that happens, it is conceivable that some bioengineer might clone a new person from a cell of the dead person. Thus, the dead person has reproductive potential, and should be considered living.

If we were discussing a less nebulous concept, the "dead people are actually living" result could be taken as a sort of reductio ad absurdum, a proof that this definition must be wrong because it leads to impossible conclusions. However, for now we will simply state that this seems a disappointing result, and that it seems profitable to search further for a definition before giving up.

Resistance to Entropy

What is the characteristic feature of life? When is a piece of matter said to be alive? When it goes on "doing something," moving, exchanging material with its environment, and so forth, and that for a much longer period than we would expect an inanimate piece of matter to "keep going" under similar circumstances. When a system that is not alive is isolated or placed in a uniform environment, all motion usually comes to a standstill very soon as a result of various kinds of friction; differences of electric or chemical potential are equalized, substances which tend to form a chemical compound do so, temperature becomes uniform by heat conduction. After that the whole system fades away into a dead, inert lump of matter. A permanent state is reached, in which no observable events occur. The physicist calls this the state of thermodynamical equilibrium, or of "maximum entropy."
—Erwin Schrodinger, What is Life?

As a physicist, Erwin Schrodinger was in a good position to appreciate the oddity of life. A few decades before his time, the second law of thermodynamics had been formulated in its statistical form by Ludwig Boltzmann. According to this "entropy" law, any closed system moves inexorably toward a state of increasing disorder. How are living things able to postpone the inevitable for so long?

Schrodinger proposed an answer to that question, but before we get to it, let us consider whether the characteristic proposed above is a sufficient definition of life. Note that the criterion of self-replication, central to the Barrow/Tipler and Von Neumann definitions, is not even mentioned. We have at this point encountered the first of a new category of definitions: those that emphasize the ongoing processes of life rather than its reproductive potential.

Barrow and Tipler have the following to say against such definitions:

But if information preserving (or increasing) reproduction is removed from the list of physiological processes, then it seems that candle flames must be considered living organisms. Flames "eat" or rather take in fuel such as candle tallow, and they "breathe" oxygen just as animals do. The oxygen and fuel are metabolized (or rather burned) in a reaction that is essentially the same as the underlying oxidation reaction that supplies humans with their energy. Flames can also grow, and if the fuel is available in various nearby localities, move from place to place. They can even "reproduce" by spreading.

Is flame alive by Schrodinger’s definition? In other words, do flames postpone the state of maximum entropy much longer than we would expect a non-living thing to? Well, there is something self-sustaining about flames, in the sense that, as the fuel burns, new fuel is continually being exposed and heated to the point where it can also burn. Schrodinger’s definition, by itself, does not appear to exclude flame from the list of living phenomena. So the definition fails to exclude an easy example of non-life.

It is worth pointing out here that the distinction between "easy" and "difficult" examples that we made earlier is not as clear-cut as we pretended. The easy examples represent the common usage of modern people in industrialized countries, provided that they have not studied too much philosophy. (Philosophy tends to warp our usage of words.) Thus, in everyday life we would not say that an automobile or a candle flame is alive. But it was not always so, and less modern people might find it natural to regard flame as a sort of living "fire spirit" because of its dynamic behavior. The point is that flame really is a little bit life-life.

Barrow and Tipler raise another objection that strikes me as less serious:

On the other hand, tardigardes are simple organisms that can be dehydrated into a powder, and which can be stored in this state for years. But if water is added, the tardigardes resume their living functions. When in the anhydrous state the tardigardes do not metabolize. Are they "dead" material during this period?

Well, according to Schrodinger’s definition, yes, they are dead; because Schrodinger talks of life in terms of its ability to keep "doing something" and the dehydrated tardigardes cannot do anything. In fact, one of the things they cannot do is reproduce! Thus, this dehydrated state is equally problematic for the reproductively-based definitions.

I believe the normal biological term for this state is "dormancy." A dormant tardigarde bears a certain obvious resemblance to a seed that has not yet been planted and has not begun to germinate. Is a seed alive? Well, not yet. But it certainly has the potential to come alive, once the right conditions occur in its surrounding environment. Is the tardigarde alive? Well, not at the moment. It was before and it can be again.

Yet one is reluctant to classify the seed or the dormant tardigarde with things like rocks or spoons, which are equally inactive. What is the difference? It is one of potential. A rock or a spoon will never "wake up" and start living. For convenience sake, then, we can speak of seeds and tardigardes as having life in a "potential" form.

There is a partial analogy here to the concept of potential energy. When you throw a baseball straight up in the air, it begins with a lot of kinetic energy—energy of motion. But it gradually slows down as it rises, finally stops, and falls at gradually increasing speeds. Except for the effects of friction, it would re-attain its original speed before reaching ground level. A physicist would say that as the ball was rising and slowing down, its kinetic energy was gradually being converted into potential energy; and when it was falling and speeding up, the potential energy was being reconverted into kinetic energy. This makes sense to physicists because they can then speak of the law of conservation of energy: the total amount of energy in a closed system never changes, it just gets transformed in various ways.

I’ve digressed here to make a minor point, which is simply that when we speak of potential life, this concept is not meant to imply anything like conservation of life. Life is not conserved. Killing something does not create new life elsewhere; it just reduces the number of living things in the world. And seeds can rot or get eaten before they germinate, so the potential for life can also be destroyed. The concept of potential life merely allows us to distinguish between the possible futures of various types of currently nonliving things.

But let us return to Schrodinger. In his book, he goes on to suggest just how it is that life is able to resist entropy:

How does the living organism avoid decay? The obvious answer is: By eating, drinking, breathing, and (in the case of plants) assimilating. The technical term is metabolism. The Greek word means change or exchange. Exchange of what? Originally the underlying idea is, no doubt, exchange of material . . . That the exchange of material should be the essential thing is absurd. Any atom of nitrogen, oxygen, sulphur, etc., is as good as any other of its kind; what could be gained by exchanging them? For a while in the past our curiosity was silenced by being told that we feed upon energy . . . Needless to say, taken literally, this is just as absurd. For an adult organism the energy content is as stationary as the material content. Since, surely, any calorie is worth as much as any other calorie, one cannot see how a mere exchange could help.

What then is that precious something contained in our food which keeps us from death? That is easily answered . . . The device by which an organism maintains itself stationary at a fairly high level of orderliness (= fairly low level of entropy) really consists in continually sucking orderliness from its environment . . . In the case of higher animals we know the kind of orderliness they feed upon well enough, viz. the extremely well-ordered state of matter in more or less complicated organic compounds, which serve them as foodstuffs. After utilizing it they return it in a very much degraded form—not entirely degraded, however, for plants can still make use of it. (These, of course, have their most powerful supply of "negative entropy" in the sunlight.)

The first time I read this, it seemed like a brilliant insight. However, problems arise when you think about it more carefully. In his notes to the chapter, Schrodinger goes on to say:

But F. Simon has very pertinently pointed out to me that my simple thermodynamical considerations cannot account for our having to feed on matter "in the extremely well ordered state of more or less complicated organic compounds" rather than on charcoal or diamond pulp . . . And so Simon is quite right in pointing out to me, as he did, that actually the energy content of our food does matter; so my mocking at the menu cards that indicate it was out of place. Energy is needed to replace not only the mechanical energy of our bodily exertions, but also the heat we continually give off to the environment. And that we give off heat is not accidental, but essential. For this is precisely the manner in which we dispose of the surplus entropy we continually produce in our physical life process.

Schrodinger does not actually go far enough here to address the objections that F. Simon raised. Schrodinger merely admits that, as well as being a source of negative entropy, food is also important as a source of energy.

But consider the example of a steam locomotive. In a sense, it feeds on coal or some other fuel. It burns the fuel, which enables it to perform work (moving forward). Like living things, the locomotive degrades the coal and exhausts entropy (heat) to the environment. But can we really say that the fuel is its source of order? The order was imposed on the locomotive in the factory where it was built, where the parts were arranged in a certain fashion. The order is maintained by engineers and repairmen who perform tasks such as cleaning, lubricating, and adjusting the mechanisms. The fuel contributes nothing to the orderliness of the machine, but simply provides it with motive power.

Isn’t it conceivable, therefore, that living things use food simply as a source of fuel, rather than a source of order per se?

Well, that wouldn’t be quite right. Food is not simply a source of fuel; it is also a source of material which the body can use to build tissues and perform repairs. But if you look at food from a standpoint of order, one of the first things you will notice is that the large-scale order in the food is destroyed before our bodies can assimilate it. First, we generally don’t eat something until after it is dead; then we chop it into pieces, cook it, and chew it up into mush before swallowing.

What is more, the digestive process also breaks down many complex organic molecules into simpler forms before they are assimilated. Thus, all the proteins are broken down into amino acids and only reassembled into proteins after they are digested. Further, the body can use protein for more than one purpose: if more is available than is needed to build tissues, the body can burn the protein for energy instead. The food we eat does not carry any mandate with it about how it should be used.

So the image of life as "sucking orderliness from its environment" is a little deceptive. Rather, the body sucks material from its environment. The body destroys most of the order in this material, producing building blocks that are then absorbed and rearranged into the body’s preexisting order.

Eating is so common that we tend to forget what an extraordinary process it is. We eat corn flakes, and yet, we do not become corn flakes. The corn flakes are changed into us. In [Richard] Feynman’s splendid phrase, "Today’s brains are yesterday’s mashed potatoes."
—I. S. Shlovskii and Carl Sagan, Intelligent Life in the Universe

However, it is true, as Schrodinger points out, that all life activities increase entropy, and that the body must exhaust this entropy back into the environment (primarily by radiating heat).

To summarize, what have we learned from Schrodinger?

Living things are systems with a characteristic order that persists over time.

Living things are active. (Even if an organism appears to be sitting still, processes are going on inside it.)

Living things are open systems that exchange material and energy with their environment.

Living things increase the entropy in the environment around them.

All these things appear to be true of life, but they do not constitute the definition we were looking for. We know these factors are not a complete set of sufficient conditions, because they apply to phenomena such as candle flames that are not normally considered alive. Besides that, we have not determined which of these criteria are minimal independent criteria for life, and which are simply consequences that can be derived from the more basic criteria.

Autopoiesis

Every five days you get a new stomach lining. You get a new liver every two months. Your skin replaces itself every six weeks. Every year, ninety-eight percent of the atoms of your body are replaced. This nonstop chemical replacement, metabolism, is a sure sign of life. This "machine" demands continual input of chemical energy and materials (food).

Chilean biologists Humberto Maturana and Francisco Varela see in metabolism the essence of something quite fundamental to life. They call it "autopoiesis." Coming from Greek roots meaning self (auto) and making (poien, as in "poetry"), autopoiesis refers to life’s continuous production of itself. Without autopoietic behavior, organic beings do not self-maintain—they are not alive.

An autopoietic entity metabolizes continuously; it perpetuates itself through chemical activity, the movement of molecules. Autopoiesis entails energy expenditure and the making of messes. Autopoiesis, indeed, is detectable by that incessant life chemistry and energy flow which is metabolism. Only cells, organisms made of cells, and biospheres made of organisms are autopoietic and can metabolize.

—Lynn Margulis and Dorion Sagan, What is Life?

The idea of metabolism was previously mentioned in Schrodinger’s definition. Thus, we were already aware that life exchanges materials with the environment. However, the idea of autopoiesis goes a little bit further than this. As Margulis and Sagan point out, it is not just certain parts of the organism that participate in the exchange of materials with the environment. Rather, every part of the body is involved in this exchange, and gradually gets replaced over time.

Can we use autopoiesis as a single sufficient criterion for identifying life forms? Well, it certainly fits the easy examples. You and I and all the animals and plants and bacteria are autopoietic. Can we use this criterion to clarify the difficult examples? Margulis and Sagan consider non-replicating examples such as mules:

Replication is not nearly as fundamental a characteristic of life as autopoiesis. Consider: the mule, offspring of a donkey and a horse, cannot "replicate." It is sterile, but it metabolizes with as much vigor as either of its parents: autopoietic, it is alive. Closer to home, humans who no longer, never could, or simply choose not to reproduce can not be relegated, by the strained tidiness of biological definition, to the realm of the nonliving. They too are alive.

Margulis and Sagan go on to consider viruses:

In our view, viruses are not [alive]. They are not autopoietic. Too small to self-maintain, they do not metabolize. Viruses do nothing until they enter an autopoietic entity: a bacterial cell, the cell of an animal, or of another life organism. Biological viruses reproduce within their hosts in the same way that digital viruses reproduce within computers. Without an autopoietic organic being, a biological virus is a mere mixture of chemicals; without a computer, the digital virus is a mere program.

Smaller than cells, viruses lack sufficient genes and proteins to maintain themselves. The smallest cells, those of the tiniest bacteria (about one ten-millionth of a meter in diameter) are the minimal autopoietic units known today. Like language, naked DNA molecules, or computer programs, viruses mutate and evolve; but, by themselves, they are at best chemical zombies. The cell is the smallest unit of life.

Now consider the case of the candle flame. A flame is composed of gases that are continuously in motion. New material enters through the bottom of the flame, and waste products such as smoke and carbon dioxide exit through the top. All the molecules in the flame are regularly replaced, yet the flame itself persists. Is the candle flame an autopoietic system? Seemingly it is. Thus it would seem that things can be autopoietic without being alive.

It also seems conceivable that something could be alive without being autopoietic, at least not in the full sense of the constant and gradual replacement of molecules and cells. Suppose that we succeed in creating artificial and mechanical life forms in the future. Such beings might require only the occasional replacement of major modules or "organs." However, we are obviously indulging in speculation at this point.

Speculation alone should not perturb our definition greatly, but the case of the candle flame does suggest that life has some additional essential feature that goes beyond autopoiesis.

Responsiveness

Why don’t we consider a candle flame to be living? Well, suppose we compare it with something that is living and see what the differences are. Let us compare the candle flame with, say, a mouse.

Both the candle flame and the mouse need fuel. However, when the candle runs out of wax, it is doomed. When the mouse runs out of food, it just starts exploring until it finds some.

You could argue that flames in general can spread to and engulf new fuel, such as when a spark starts a forest fire. But there are limitations to this spreading, such as that fire outdoors will tend to spread downwind. It may be that better sources of fuel are lying upwind, and so never get used. A mouse, by contrast, can move upwind, or uphill. It can see or smell food at a distance, or it can just keep searching around until it finds something.

Note that this behavior is possible because the mouse doesn’t have to feed at every moment, as the flame does. It can store up energy inside and use it to survive until it finds more food.

Now, besides the risk of running out of fuel, there are other hazards that can affect the candle flame and the mouse. I can walk up to the candle flame and blow it out. But if I decide to kill the mouse, I will find it a little more difficult. Most likely, it will see me coming, and run for cover.

So the mouse is much better than the flame both at finding new fuel and evading dangers. Both of these factors involve the relationship between the mouse and its environment. This type of observation must have inspired Herbert Spencer’s definition of life in Principles of Biology as "the continuous adjustment of internal relations to external relations."

Barrow and Tipler quote this definition and disapprove of it, on the grounds that "such definitions possess rather extreme ambiguities." Taken by itself, Spencer’s phrase is certainly insufficient, because it doesn’t make it clear what kind of adjustment is taking place. Does he mean that the parts of the body actually try to mimic the arrangement of items in the outside environment? It doesn’t seem like there would be much point in such a behavior. Still, there is something promising about Spencer’s definition, because it focuses on the relationship between an entity and its environment; and, as we have seen, this is an area where the mouse and the candle flame differ greatly.

In the book Doubt and Certainty in Science, the British biologist J. Z. Young compares living things to rivers, because they have a pattern that persists even though new matter is constantly flowing throw them. Young then goes on to say

But the [living] organization is vastly more complicated than that of any river. It is kept in certain channels by the environment, acting in a sense as do the banks. If a stream stops, the banks remain, and therefore a river that has dried up may form again the same patterns. But the living patterns are so complicated that they are kept intact only by their continued activity. If they stop they are never restarted. The living patterns have developed a wonderful permanence none the less. They have the characteristic that every time there is any change in the banks the swirls make a compensating change and thus keep intact.

Young abandons the metaphor at this point because it is difficult to go any further with it. But he has already added something that goes beyond Spencer’s statement. The key phrase is "the swirls make a compensating change and thus keep intact." Of course, when you think about it, this is a fairly obvious observation. The mouse does not run away from food, nor does it run toward enemies. Staying alive is a serious business; it doesn’t happen by accident.

All this suggests a new definition of life, which I shall formulate as follows:

Living things are systems that tend to respond to changes in their environment in such a way as to promote their own continuation.

When I say "new," of course, I mean only "new to me." Such a simple and obvious definition has probably been proposed by someone before. Yet in my reading on the subject, I have never run across it.

Of course, one can imagine some objections to this definition. Someone will pop up and say, "What about the tardigardes when they have been dehydrated? They are not responsive to anything." This is true, but as I have argued previously, they are also not living in anything but a potential sense.

Of more serious concern are the cases where life acts against its own continuance. The simplest example is that of the man who commits suicide. Now, it has been said that a failed suicide attempt is actually a "call for help," and this may be so. But many of the suicidal adopt methods that are reliable, and which they must have known in advance would really kill them. So the objection here is, that the person who commits suicide was indeed alive, but they were not acting in such a way as to stay alive. This seems to be a counterexample to our definition.

Yet many or most cases of suicide result from mental illness. Such cases are comparable to any case where something has become broken and is no longer able to function according to its original nature. You don’t define the function of an automobile on the basis of the way it behaves when it has a thrown rod or a broken head gasket.

Other cases of suicide may be quite rational, most notably among elderly people with painful and lingering terminal diseases. But in that case you see that something is still "broken." In this case, the mind is sound but the body is not functioning properly.

Then, of course, there are people who take needless risks, such as smoking cigarettes or driving without seatbelts. These actions do not promote survival, but the risks are long-term and difficult to comprehend vividly. Even people who smoke are not stupid enough to walk in front of a speeding train. For all such people, we can conclude that the balance of their actions tends to promote survival. But we can still find in these examples an important addendum for our definition:

Living things are not perfect. Through internal breakdowns, mistakes, insufficient skill or strength, or sheer bad luck, all living processes fail eventually and die.

There is another important lesson to be drawn. Although the balance of life activities are such as to promote survival, it does not follow that survival is the goal of life activities. In other words, animals eat because they are hungry, not because they want to survive. But the instinct to eat when hungry is one of many instincts that tend, on balance, to promote survival.

The distinction may not seem important when we are talking about animals, but among humans it is crucial. We are able to understand the long-term impact that many of our actions will have on our survival. Yet our ability to understand such things is a relatively recent acquisition, and tends to get bypassed by more primitive instincts. Thus for example, we tend to eat more and exercise less than is good for our health. Presumably in ancient times food supply was irregular and incitements to physical activity were plentiful, so our instincts were suitable for those times. Our intellectual understanding of this fact does not always provide sufficient motivation to change our behavior today.

Thus, in general human beings are capable of doing things which we know do not promote our chances for survival. Yet the balance of our activities must be such as to promote survival, or in fact we would not last very long.

Of course, many of our activities have no obvious bearing on survival at all. For instance, we come home from work and we decide to watch Star Trek or Seinfeld, or read a book or do a crossword puzzle. These activities do not fulfill any immediate physical need. But note that we pursue these activities within certain boundaries. We do not, as a rule, amuse ourselves by running in front of cars on the freeway, or by hitting ourselves on the head with a hammer. And our leisure activities presumably are driven by instincts that have some general survival value, such as the urge to learn, to solve problems, or to form social relationships.

So our proposed definition of life is consistent with a variety of observations about living things. But seemingly it must collapse under the weight of a significant counterexample, which is the fact that in nature, many things willingly sacrifice their own lives for the sake of others. Let us go on to explore this theme in a little more depth.

Continuance Through Others

Let us begin by listing some of the situations in which living things have been known to sacrifice their lives for others:

Mothers defend their children against attackers or potential attackers. For instance, a bird will sometimes harass cats to keep them away from her nest.

People often attempt to rescue even total strangers from burning cars or other dangerous situations.

Honey bees fight intruders, such as bears, that threaten their hive. Whenever a bee stings a bear, the bee dies from losing its stinger.

Soldiers or terrorists will sometimes agree to "suicide missions." An example is the Japanese fighter pilots who would attempt to crash their planes into American naval ships in WWII. Similarly, terrorists will sometimes hide a bomb under their clothing, go into a crowded public place, and detonate themselves.

These examples fall into two classes. In the first two, the living thing does something risky, but in the second two, the living thing does something certain to lead to death. The first and less extreme class is a lot more common than the second.

Let us examine the case of mothers defending their children first. What is the relationship of mothers and their children? Aristotle has this to say:

For this is the most natural of the functions of such living creatures as are complete and not mutilated and do not have spontaneous generation, namely to make another thing like themselves, an animal an animal, a plant a plant, so that in the way that they can they may partake in the eternal and the divine. For all creatures desire this and for the sake of this do whatever they do in accordance with their nature . . . Now the living creature cannot have a share in the eternal and the divine by continuity, since none of the mortal things admits of persistence as numerically one and the same, but in the way that each creature can participate in this, in that way it does have a share in it, some more some less, and persists not as itself but as something like itself, not numerically one, but one in species.
—De Anima II.4, trans. Hugh Lawson-Tancred

In other words, because the individual cannot survive forever, it creates offspring similar to itself so as to survive through them. Due to the vagaries of sexual reproduction, the offspring resemble each parent only in certain respects. Thus, survival through one’s children is only a partial and limited sort of survival. Yet all individuals die, so this partial survival through children is the best backup plan available.

Today, Aristotle’s explanation of reproduction may seem a bit too purposive. He credits all living things with an abstract desire to "partake in the eternal and the divine." And he makes it sound as if living things all understand their own mortality and thus deliberately choose to reproduce because of this. I. S. Shlovskii and Carl Sagan provide a more modern spin to the importance of reproduction:

Can it be that reproduction is in some sense the "point" of biological activity? We can imagine an organism which carries out metabolism and all the other functions ordinarily ascribed to living systems in elementary biology textbooks; which has very efficient repair mechanisms, so that it easily survives the vicissitudes of its environment; and which has no reproductive organs and never reproduces. We can imagine such an organism, but we can never find one. Why not? Because there is no way for such an organism to arise. The only mechanism which we know for the production of biological complexity is evolution by natural selection, the differential survival of organisms which, by chance, are best adapted to their environments. But natural selection can occur only if the well-adapted organisms reproduce themselves. Thus, the development of complexity in living systems is intimately connected with self-replication.
—Intelligent Life in the Universe

Evolution can provide organisms with reproductive instincts even if the organisms themselves do not understand that they will die, and that the partial continuation through offspring is their only long term chance for survival.

It makes perfect sense to regard reproduction as one application of a more basic characteristic of life, which is its tendency toward self-preservation. By looking at reproduction in this way, we avoid the problems with the reproductively-based definitions of life that we examined earlier.

Let us return to the example of the mother risking her life to protect her children. We can now see that the mother’s actions follow naturally from our definition of life. By preserving her children, she is in some sense preserving herself as well.

Now consider the honey bee, which gives its life by stinging a bear that is attacking the hive. The case is different here, for a worker bee can never have any offspring. Only the queen and a small number of drones play any role in reproduction. In this organization, the worker’s only chance of long-term continuance is to protect the queens and drones. It also needs to protect its fellow workers so that they, in turn, can help protect the queens and drones; and it needs to protect the supply of honey in the hive that must be there to enable them all to survive the winter. So in a general way the bee must defend the whole hive to ensure its own continuance, and thus it makes sense for it to give its life by stinging the bear.

The examples of suicide missions and rescuing strangers are far less straightforward. The sociobiologists, such as Edmund O. Wilson, have proposed that altruism can be a genetically motivated trait, even in humans. However, we can also see that social forces can and do encourage altruistic behavior in all of us. Whether the causes are genetic or cultural, the underlying logic is much the same. Human beings are generally social creatures and our individual survival depends on our being part of a viable social group. Actions that benefit other members of the group can thus have an indirect future benefit either for the actor, or for relatives or descendants of the actor, or at the very least for members of the same species. As Aristotle says, a living thing that gives its life for others of its kind "persists not as itself but as something like itself, not numerically one, but one in species."

Homeostasis

Following J. Z. Young, we have stated that living things are organizations or systems. In other words, although the individual constituents are gradually replaced, their overall arrangement stays roughly the same or changes only gradually. This maintenance of a constant state is referred to as homeostasis.

This is an interesting point because it causes us to look at life from the inside. During our previous comparison of candle flames and mice, we focused on the externals of their behavior, on their relationship with the environment. And this was appropriate to a degree. If you tried to define life strictly in terms of the relationships of its internal parts, you would be missing something essential. For none of the internal processes would be possible unless the organism succeeded in obtaining food, evading predators, and so on.

Yet the threats to an organism’s existence do not all come from the outside, and even the ones that start on the outside do not always stay there. Disease organisms, whether viruses or bacteria, often invade the system. In higher organisms, the immune system exists to battle such invaders. The immune system also attacks the body’s own cells if they become cancerous.

Aside from handling such outright threats, the body must coordinate a lot of complex internal processes during the course of its day to day functioning. This is true even of single-celled life forms. Hans Kupper points out that

Even in a simple bacterial cell there are estimated to be around a million functional molecules of two to three thousand different kinds. Each of these molecules carries out a particular, specialized task, which in general is indispensable for the maintenance of the functions of life.

Coordinating these functions involves the use of feedback. In other words, the internal processes are regulated according to need. A thermostat is a simple example of a feedback mechanism: by turning the heater on or off, it affects the temperature, and the changing temperature in turn affects the thermostat. The thermostat reacts in such a way as to minimize the temperature deviations, and for this reason is said to be using negative feedback. Negative feedback is a key element by which the organism coordinates its internal systems and promotes homeostasis. Fritjof Capra gives a nice introduction to feedback in his book The Web of Life.

Do we need to incorporate the concept of feedback into our definition of life? Well, it is important to recognize that feedback per se does not necessarily promote life. Positive feedback can increase the fluctuation in a system and cause it to self-destruct. Even negative feedback can be inappropriate, if it preserves the stability of some subsystem at the expense of the system as a whole. The central point seems to be that the ensemble of inner processes tends toward homeostasis, the maintenance of the overall pattern of the system.

Remember that we are searching for the minimal criteria to identify life. From this point of view, it is sufficient to amend our definition by adding one simple clause:

Living things are systems that tend to respond to changes in their environment, and inside themselves, in such a way as to promote their own continuation.

We shall now go on to examine some of the concepts used in this definition in a bit more detail.

Patterns of Complexity

Biology, like physics, has ceased to be materialist. Its basic unit is a non-material entity, namely an organization.
—J. Z. Young, Doubt and Certainty in Science

Young’s statement follows from the fact of metabolism, the constant and gradual replacement of every molecule in your body. Since the material flows in and out of the organism, the organism cannot be considered to be a certain blob of matter, but is instead a characteristic pattern in which matter is organized.

It happens that the nature of pattern and complexity is a topic that has recently received a good deal of study. Following are some of the key results for the study of life. 

Algorithmic Complexity

Starting in the 1960’s, researchers including Gregory Chaitin, R. J. Solomonoff, and A. N. Kolmogorov created the branch of mathematics known as algorithmic information theory (AIT). This theory defines patterns in terms of computation and computability. (A number of articles on AIT can be found on Gregory Chaitin’s homepage at http://www.cs.auckland.ac.nz/CDMTCS/chaitin/.)

For simplicity’s sake, you can discuss all the issues relating to patterns in terms of binary strings, that is, strings of zeroes and ones. This is the way that all information is stored in computers. Even if something is three-dimensional, multicolored, and so on, there is always a way of expressing that collection of properties as a long binary string. We will refer to this as the representation string for that physical object.

Of course, we are not able to describe any physical system in complete detail. For example, at the level of subatomic particles we run into quantum measurement problems. In practice, when you create the representation string, you must choose some particular level of detail and accuracy. This procedure is known as coarse graining.

Within AIT, a pattern is defined simply as a program for generating a particular binary string. A program is a collection of steps that can be executed on a computer.

Note that every program on a computer is itself stored as a binary string. Within AIT, the complexity of any given string is defined simply as the length of the shortest program that will generate that string. This program is referred to as the minimal program for that string. (The literature seems to imply that there could be more than one minimal program—programs that are different, but of equal length.)

Actually, the length of the minimal program will vary to some extent, depending on the computer and the programming language that are used. Thus, when comparing the complexity of different strings, it helps to define their complexity in terms of the same computer system. (For very long strings, the differences between computer systems become less and less significant.)

A particular string is said to be random if the minimal program for generating that string is about as long, or longer, than the string itself. Thus, randomness is also referred to as noncompressibility; the inability to recreate something from anything shorter than itself.

Actually, the concept of algorithmic complexity does not correspond very well to the concept of complexity in everyday life. You can see this because algorithmic complexity is maximal for completely random strings. In everyday life, we draw a distinction between randomness and complexity. Complexity does involve some kind of order, though the order is not of a simple kind.

Logical Depth

The scientist Charles Bennett, of IBM, has suggested the concept of logical depth, as a measure that more closely resembles our everyday idea of complexity. The logical depth of a string is defined as the number of runtime steps it takes for the minimal program to create that string. (This number can differ from the length of a program because a program can include loops or subroutines that are executed repeatedly, and branches that selectively skip statements.)

Rudy Rucker, in his book Mind Tools, says the following about logical depth:

Bennett and Chaitin, who are colleagues, speak of the two extremes of complexity as crystal and gas. The atoms of a crystal have the property that they are very obviously arranged according to a simple rule. They are like soldiers on a parade ground. The atoms of a gas have the property that they are totally disordered, and are not arranged according to any rule much shorter than an actual listing of the atom’s positions. Patterns that we find interesting — things such as living organisms and manmade artifacts — lie midway between the extremes of crystal and gas. One of the reasons Bennett invented the notion of logical depth is that he wanted depth(gas) and depth(crystal) to be small, but depth(organism) to be high.

And

Bennett . . . argues that it may be appropriate to characterize living organisms as physical structures that code up as bit strings with depths much larger than their lengths.

Now, logical depth could be a property of organisms. But I’m not sure how you would go about proving it. And it seems clear that logical depth is a property of at least some things that are not organisms. Consider the Mandelbrot set, a favorite of computer hobbyists who like to play with fractal mathematics. The Mandlebrot set is based on a simple function that, when applied repeatedly, generates some particular value for each point in a two-dimensional plane. In pictures of the Mandlebrot set, different colors are assigned to different ranges of values. Now imagine a physical object: a large and very beautiful poster representing some portion of the Mandlebrot set. The poster coloring can be described by a very simple program that repeats a function repeatedly for each point on the poster. In other words, the poster has great logical depth. But no one would mistake it for a living thing.

Perhaps this seems too abstract, because a poster is flat and we’re just describing its coloring. The Mandlebrot set can be represented in a three-dimensional way, so that it looks something like a relief map with mountains and valleys. The height of each point conveys the same information as the colors in the poster. So then you have a three-dimensional object with tremendous logical depth. But it doesn’t perform any of the functions we have associated with life: it doesn’t metabolize, it doesn’t respond to the environment. It would not be classified as alive by any of the definitions we have considered, not even the most naïve ones. 

Effective Complexity

In his book The Quark and the Jaguar, the physicist Murray Gell-Mann discusses another type of complexity that he calls effective complexity. This concept, like the concept of logical depth, is an attempt to supplement AIT with a concept closer to our intuitive sense of what is complex.

Gell-Mann argues that in real life, we instinctively abstract out of our experience the elements that are orderly. We use these orderly elements to develop and test mental schema, which are models of selected elements of our experience. Gell-Mann defines the effective complexity of a phenomenon as the length of the schema that describes the regularities in that phenomenon.

For example, consider the following binary string:

1011101011001001011...

In this string, every third digit is "1":

1..1..1..1..1..1..1...

The remaining digits have no obvious order:

.01.10.01.00.00.01....

A schema for this string could describe only the regular component: the recurrence of 1 at every third digit. As this is a very simple schema, the effective complexity of this string is quite low. The algorithmic complexity is much higher because the minimal program has to be able to create the complete string, rather than just the portions that have obvious regularities.

Ranges of Fluctuation

Let us step back and ask ourselves how these AIT concepts apply to living things. If a living thing is an organization, then it must be possible to construct either a pattern that gives create a complete description of the organism, or a schema that gives a complete description of the regular elements in the organism. Further, that program or schema must be shorter than the representation string for the organism.

Does such a program or schema exist? At first glance, it might seem that the DNA of an organism plays such a role. It certainly is reasonable to describe DNA as a program, which is executed by the cellular "computer" consisting of RNA and proteins. It includes a plan for building the organism from scratch (morphogenesis) and instructions for the behavior of each type of cell.

However, the organism is shaped and modified by environmental influences, and the DNA does not record these. For example, as we learn and experience the world, our brain structure changes through the growth of dendrites and so on. Similarly, the formation of the body in childhood can be affected by the availability of various nutrients. More disturbing, perhaps, is the following observation:

An average human is normally host to billions of symbiotic organisms belonging to perhaps a thousand different species . . . His phenotype is not determined by his human genes alone but also by the genes of all the symbionts he happens to be infected with. The symbiont species an individual carries usually have a very varied provenance, with only a few being likely to have come from his parents.
—Juan Delius, "The Nature of Culture," in M. S. Dawkins, T. R. Halliday, and R. Dawkins, eds., The Tinbergen Legacy; quoted in Daniel C. Dennett, Darwin’s Dangerous Idea

Further, the body is changing all the time. On the microscopic level, the actual constituents of the body are constantly changing through metabolism. On the macroscopic level, the overall shape of the body changes every time we perform a voluntary or involuntary movement (walking, breathing, etc.). What is more significant, the changes that the organism goes through are often in response to environmental conditions that cannot be predicted in advance.

Thus, if you regard DNA as a pattern of the organism, you must be using a representation string that was developed at an extreme level of coarse graining. It would actually be a level much coarser than our everyday perception of living things, which includes an awareness of features of organisms, such as characteristic behaviors, that are strongly shaped by the environment.

To regard DNA as a schema for living things is equally unsatisfactory. A schema describes the regularities in a phenomenon, but as we have seen, an organism can acquire regularities through experience. This may seem paradoxical, but an acquired trait needs to be considered a regularity because it persists and affects all future behavior, as well as affecting the acquiring of new traits.

An additional problem arises because schemata describe regularities, yet organisms do not seem to have any characteristics that are perfectly regular. Certainly the relative positions of the limbs change all the time. Aside from that, a property such as temperature fluctuates around an average. Our height also varies on a daily basis, due to compression of the soft parts of the spine during the day. The ears and nose continue gradually growing during adulthood. Bones seem permanent, but the calcium in bones can be temporarily shifted to other parts of the body, or the bones can gradually waste away, resulting in osteoporosis. Our eyes may be bloodshot one day and white the next. Our DNA itself frequently develops errors, and various error-correction mechanisms exist within the cell to compensate for this fact.

Paul Weiss formalized the idea of a system is in the following way:

Weiss now denotes the complex S as a system if the fluctuations in the properties of the whole complex are significantly smaller than the sum of the fluctuations for the subsystems.
—Hans Kupper, Information and the Origin of Life

This gives us an interesting way of thinking about living systems. All the apparent regularities in an organism appear really to be fluctuations around average values. In addition to this, the average values themselves gradually change over time as the organism passes from infancy into adulthood and decays into old age.

To describe the order in an organism, then, you would need to give a list of properties and explain the characteristic ways in which each property varies: the average value, and the range, rate, and cause of variation, among other things.

Such a description could be considered a schema of a sort, but it has a certain mushy quality to it. As a result, to recognize a particular organism, you must make use of mechanisms with mushy or gradual constraints. The disciplines of fuzzy logic and neural network theory have developed examples of such mechanisms.

Behavior

We have been considering what it means to say that living things have some characteristic order, or pattern of organization. The most useful concept we have found for describing that order is Paul Weiss’s notion of a system as a collection of properties with fluctuating values.

But this is only one aspect of life. We have next to explore what it means when we say that a living thing responds to changes in its environment, in such a way as to promote its own continuation. We can begin by asking whether this property, behavior, is really a universal property of life. S. E. Luria, Stephen Jay Gould, and Sam Singer have this to say in their book A View of Life:

In the broadest sense of the term, behavior consists of everything an organism does as it goes about the business of survival and reproduction, including reflex activities that help maintain the constancy of its internal environment. It is therefore not surprising that animals, especially vertebrates, with their elaborate brains and active life-styles, engage in the most complex behavior of many organisms. But behavior is by no means limited to animals with complex brains, or even to animals.

Unicellular organisms, both prokaryotic and eukaryotic, are capable of several kinds of behavioral responses to specific stimuli. For example, some species of bacteria that contain particles of iron oxides in their cytoplasm preferentially orient themselves in certain directions when places in weak magnetic fields. Whether or not this simple behavior has any adaptive significance remains to be discovered . . . E. coli and some other bacteria react to certain chemicals in their environments by changing the direction in which their flagella rotate. This response occurs because these bacteria have some kid of processing mechanism in their cell membranes that causes them to swim towards nourishing chemicals and away from noxious ones.

Protists, with their elaborate unicellular body plans, have a behavioral repertoire that is correspondingly more complex than that of prokaryotes. An amoeba, for example, can detect the presence of edible cells and particles in its immediate vicinity and, as long as the intended meal does not move too far out of range, will relentlessly pursue it by extending its pseudopods. Amoebas also show specific behavioral responses when exposed to light. Bright light causes an advancing pseudopod to be withdrawn. And when an entire amoeba is suddenly exposed to bright light, it draws in all of its pseudopodia and vigorously contracts to form a spherical blob of protoplasm.

As a group, plants are among the least motile organisms and their behavior is correspondingly less complex than that of creates that actively move through the environment in search of food, shelter, and mates. Nevertheless, many plants do engage in simple, predictable behaviors. As discussed in the previous chapters, most plants react to light by bending toward it, and roots grow toward the pull of the earth’s gravitational field, while stems preferentially grow away from it. Some plants move their leaves in response to touch. This is true of the Venus fly trap, which rapidly closes a pair of hinged leaves to trap insects, and of the "sensitive plant" (Mimosa pudica), which upon being touched, heated, or stimulated by an electrode rapidly folds up its leaves and leaflets.

But what is behavior, in general? What do all these examples of specific behaviors have in common? Let us for the moment be a bit less general than mssrs. Luria, Gould, and Singer, and restrict ourselves to the interactions between an organism and its external environment. There seem to be two factors in behavior: input and output. The organism is, as it were, struck by something from outside of itself, and as a result of this, the organism moves.

But a billiard ball also is struck by something outside of itself, and also moves as a result. Yet we do not say that it responds to its environment. The motions of living things have the following features:

• Motion often involves the expenditure of more energy than was contained in the input. Aristotle intimates something similar in De Motu Animalium, Chapter 7: "It is not difficult to see that a small change occurring in an origin sets up great and numerous differences at a distance—just as, if the rudder shifts a hair’s breadth, the shift in the prow is considerable."

For example, suppose I see a tiger and flee. In this case, a small amount of light struck my retina. The energy supplied was not enough to move my body significantly. Yet my body did move and it took a great deal of energy to make it move. The necessary energy was supplied from within, by converting stored chemical energy into kinetic energy.

Similarly, a plant struck by light from a certain direction, will tend to expend its own resources to send out new growth in that direction.

• Motion does not necessarily continue in the same direction as the input, nor is it necessarily modified in the same direction is the input. In other words, you cannot simply calculate the subsequent movement of the organism by adding vectors that describe the momentum of the input to vectors that describe the current momentum of the organism.

Actually, the laws of physics still apply to organisms. If you shoot a man out of a cannon, you can predict his path in much the same way as you do that of a cannon ball. But the majority of our movements are not the result of such massively energetic inputs, and so the initial effect of the impact is overwhelmed by the body’s internally generated response.

• Motion may result from input after some considerable delay. For example, if a man pinches a woman, she may slap him a moment later, or she may file a lawsuit against him a year later.
• Motion is made possible by internal variations in the organism that tend to remain within typical ranges of fluctuation. For example, muscles tighten or relax to cause the movements of limbs.
• Different species exposed to similar input may move in different ways.
• Within a species, different individuals exposed to similar input may move in different ways.
• The same individual, when exposed to similar input on different occasions, may move in different ways.
• An organism, on encountering an obstacle to motion, may go around the obstacle and continue motion in the original direction.

To summarize, the relationship between the input and the response is an indirect one. The points listed above are various types of indirection. Such examples could probably be multiplied.

The causal relationship between input and output is a complex one because of things that happen inside the organism. Internal processes mediate between the input and the response, and as a general term I shall refer to such processes as mediation. Mediation can include, but is not limited to, the phenomenon of intelligence. For instance, reflex reactions are also the result of internal mediation, though it is of a primitive sort.

Indeed, it could be argued that inputs are not the causes of our actions at all, since as we have seen, no particular response follows necessarily from any particular input. Further, you can be sitting thinking, undisturbed by the conditions around you, and then suddenly reach a conclusion that causes you to take some action. Aristotle again, in De Motu Animalium:

For the animal moves and progresses in virtue of desire or choice, when some alteration has taken place in accordance with sense-perception or phantasia. . .

This, then, is the way that animals are impelled to move and act: the proximate reason for movement is desire, and this comes to be either through sense-perception or through phantasia and thought . . .

For sense-perceptions are at once a kind of alteration and phantasia and thinking have the power of the actual things. For it turns out that the form conceived of the [warm or cold or] pleasant or fearful is like the actual thing itself. That is why we shudder and are frightened just thinking of something.

The term phantasia in the passage is apparently an untranslated Greek word for imagination.

Now, all this is true but it does not follow that our actions are uncaused, or even uncaused by input from the outside. For even in the case where we are sitting thinking, and suddenly decide to do something, the mind that is doing the thinking has been progressively modified by many inputs over the years. What does follow is that these inputs are not the complete causes of action. You put a key in a lock to open the door, but the key could have no effect unless the lock were shaped to receive it. The door opening is caused not only by the key turning, but also by all the forces that shaped the lock that receives it. Our responses are caused not just by some stimulus, but also by all the factors that have shaped us, both genetic and environmental.

Men think themselves free inasmuch as they are conscious of their volitions and desires, and never even dream, in their ignorance, of the causes which have disposed them so to wish and desire.
—Benedict de Spinoza, The Ethics, Appendix to Part I, trans. R. H. M. Elwes

While the considerations in this section are helpful secondary criteria for identifying life, they are not sufficient, either individually or in combination, to finally show that something is alive. For we might be able to create some mechanism that responds to inputs in some sophisticated and indirect way. But if its responses are not such as to promote survival, such as by avoiding threats and seeking out food, then it is not really behaving in the manner characteristic of life.

A Matter of Degree

Why are there "difficult" examples to address in definitions of life? Take the example of viruses. We have seen, so far, that Barrow and Tipler regard the virus as alive, whereas Poundstone and Margulis and Sagan regard it as not alive. Luria, Gould, and Singer offer the following conclusion in their book A View of Life:

Are viruses alive? This question is more difficult to answer because it depends on a definition of life. Suppose our definition includes the idea that living things are able to reproduce. A dog is obviously alive and is made up of living cells, but a spayed dog cannot reproduce and its genetic information dies with it; yet is alive. We may, on the other hand, define life as the possession of specific genetic information capable of functioning in living cells. Then the cells of the spayed dog are clearly alive, and so are viruses, which can multiply in living cells. Viruses reproduce and evolve if they have suitable host cells available. Are viruses any different from animals or plants, which also require specific external conditions to propagate their species? To the biologist, a virus is alive because it participates in the adventure of biological evolution.

The last sentence seems to imply a consensus, at least among biologists; but we have seen that so eminent a biologist as Lynn Margulis regards viruses as non-living. Now consider the following:

Nature proceeds little by little from things lifeless to animal life in such a way that it is impossible to determine the exact line of demarcation. —Aristotle, The History of Animals, viii: 1. Cited in Lynn Margulis and Dorion Sagan, What is Life?

Could it be that the property of life is something that can be possessed in varying degrees? Actually, this notion follows fairly naturally from the definition of life that we proposed earlier:

Living things are systems that tend to respond to changes in their environment, and inside themselves, in such a way as to promote their own continuation.

It seems clear that different systems could display varying degrees of this tendency. But do they actually? Well, we have seen that flames have a very limited ability to sustain themselves. Viruses have the ability to reproduce, but little else. In cells you find this reproductive ability as well as more advanced techniques for maintaining the integrity of the individual, such as by storing energy from sunlight or swimming toward edible foods.

However, if you visualize a graph of beings that display various degrees of life, it becomes clear that there are large gaps in the graph, rather than the smooth continuous increase that Aristotle mentions. From a flame to a virus is a large jump, and from a virus to a cell is also a large one.

Von Neumann sometimes spoke of a "complexity barrier." This was the imaginary border separating simple systems from complex systems. A simple system can give rise to systems of less complexity only. In contrast, a sufficiently complex system can create systems more complex than itself . . . (Exact) self-reproduction is a feature of systems right on the complexity barrier—systems that preserve but do not increase their level of complexity in their offspring.
—William Poundstone, The Recursive Universe

Von Neumann’s complexity barrier seems to describe the jump from a flame to a virus, which reproduces exactly. Note that no living thing currently reproduces systems more complex than itself.

The presence of these quantum leaps in survival ability does not contradict the idea that life is a matter of degree. It simply imposes a certain necessary granularity in the spectrum of living things.

If you view life as a matter of degree, then it might be possible for some organisms to exist that possess this quality in a greater degree than any we have observed so far. These might be organisms that originated on other planets, or that evolve on our planet in the future, or that are created by human efforts.

Assuming that life is a matter of degree, how might we measure the degree of life in something? Following are some possible considerations:

1. Since we have said that living things are systems characterized by some pattern or schema, we can ask whether patterns are something that can be possessed in varying degree. A useful concept in this regard is the following:

   The intensity of a process relative to a given entity . . . is a measure of how much the process simplifies the entity — how strongly the process is a pattern in the entity. It has to do with the ratio of the complexity of the process to the complexity of the entity.
   —Ben Goertzel, Chaotic Logic

   The quality of life in a system could be proportional to the intensity of the pattern or schema that defines the order in that entity.

2. You could look at the percentage of variation that occurs in the system on a daily basis. How many properties are there that vary, and over how wide a range do they vary? Assuming that this variation can be quantified, how does it compare with the elements that remain constant? The quality of life could be proportional to the ratio of the constant elements to the varying elements.
3. You could look at the percentage of variation that occurs over the course of a lifetime. In large organisms, there is a huge amount of variation since growth begins from a tiny egg and proceeds to gradually build something much greater. Yet it seems that the basic pattern of the organism emerges fairly soon, albeit on a smaller scale and with different proportions than characterize the adult form. More significant changes occur in the insect world, where you find such things as a caterpillar changing into a butterfly. The quality of life could be proportional to the degree of constancy in an organism over the course of an entire lifetime.
4. Since organisms live for widely varying lengths of time, you could say that the quality of life is proportional to the average lifespan of individuals in a species.
5. Since organisms always die eventually, and have continuance beyond death primarily only through offspring, you could focus on the differences in the ways that various organisms reproduce. The quality of life might be proportional to the degree of resemblance between a parent and its offspring. A virus reproduces more or less exactly, as do bacteria. Most higher organisms reproduce sexually, as a result of which none of the offspring exactly resembles either parent. So this criterion would have the unfortunate effect of establishing that a virus is more alive than you and me.
6. You could modify the previous criterion to reflect, additionally, the number of offspring.
7. On a species level, you could focus on how long the species persists, or on how many other species are descended from it. Is the species the origin of a major branch of the evolutionary "tree," or is it an old dead twig?

What these criteria have in common is that they each measure various ways in which patterns persist. All these criteria are interesting and perhaps worthy of having some type of name associated with them. But only the first two correlate well with our everyday use of the term life.