Cheaper than printing it out: buy the paperback book.

Out of Control
Chapter 21: RISING FLOW

Caveats aside, I discern about seven large trends or directions emerging from the ceaseless, hourly toil of organic evolution. These trends, as far as anyone can tell, are also the seven trends that will bias artificial evolution when it goes marathon; they may be said to be the Trends of Hyperevolution: Irreversibility, Increasing Complexity, Increasing Diversity, Increasing Numbers of Individuals, Increasing Specialization, Increasing Codependency, Increasing Evolvability.

Irreversibility. Evolution doesn't back up. (Also known as Dollo's Law.) There are exceptions to the no-backup principle. A whale in one sense backed up to be a fish again. But it is the exception that proves the rule. In general, current manifestations of life do not work on invading past niches.

Nor are hard-won attributes easily given up. It is an axiom in cultural evolution that technologies once invented are never uninvented. Once a vivisystem discovers language or memory it does not retreat from it.

The presence of life also does not retreat. I am aware of no geological domain that organic life has infiltrated and then retreated from. Once life settles in an environment (hot springs, alpine rock, robots) it will tenaciously maintain some presence there. Life exploits the inorganic world, recklessly transforming it into the organic. "Atoms, once drawn into the torrent of living matter, do not readily leave it," writes Vernadsky.

Prelife Earth was, by definition, a sterile planet. It is commonly accepted that although sterile, the Earth was simmering with the ingredients life needed. In essence it was a global agar plate waiting to be inoculated. Think of a immense 8,000-mile-wide bowl of pasteurized chicken broth. One day you drop a cell into it, and the next day, by the power of exponential growth, the oceanic bowl is thick with cells. In a few decades, all varieties of cells have wormed their way into every nook. Even if it took a hundred years, that is but a nano-blink in geological time. Life is born. Blink. Life is irrepressible.

Having infiltrated computers, artificial life will henceforth never retreat from being in some computer, somewhere.

Increasing Complexity. When I ask friends if evolution has a direction the common answer I get (if I get any at all) is "towards more complexity."

While it seems obvious to almost everyone that evolution moves toward greater complexity, we have few definitions of complexity that really mean anything. Modern biologists question the notion that life heads toward complexity. Steven Jay Gould has told me flatly, "The illusion of a move toward increasing complexity is an artifact. You need to build simple things first, so naturally complex things come later."

But there are plenty of simple things nature has never made. If there was not a drive toward complexity, why not stop at bacteria and invent millions of more one-celled varieties. Or why not stop at fish and fill in all possible fish forms? Why make things more complicated? For that matter, why did life start out simple? There is no law we know of that says things have to get more complex.

If there is a true trend toward complexity, there must be something pushing it. In the last hundred years a number of theories have been proposed as to what drives apparent complexity. They could be listed by the following overlapping summaries (and the year they were first postulated):

  • Runaway replication and duplication of parts makes complexity (1871).

  • The ruggedness of real environments causes differentiation of parts, which aggregate into complexity (1890).

  • Complexity is more thermodynamically efficient (1960).

  • Complexity is an inadvertent by-product of selection for other characteristics (1960).

  • A complex organism creates a niche for more complexity around it; thus complexity is a positive feedback loop amplifying itself (1969).

  • Since it is easier for a system to add a part than to remove a part, complexity accumulates (1976).

  • Nonequilibrial systems accumulate complexity when they dissipate entropy, or wasted heat (1972).

  • Chance alone produces complexity (1986).

  • Endless arms races escalate complexity (1986).

Because the term complexity is vague and unscientific at present, no one has done a systematic study of the fossil record to determine whether or not quantitative complexity increases over time. A few studies of particular short lineages of organisms have been done (using differing measures of complexity) and they have shown that sometimes some aspects of these creatures increase in complexity and sometimes they don't. In brief, we don't know for sure what happens as organisms apparently complexify.

Increasing Diversity. This one needs some careful clarification. One famous bed of fossils, the soft-bodied animals in the Burgess Shale, is currently forcing a rethinking of what we mean by diversity. As Gould tells in Wonderful Life, the Burgess Shale show a remarkable range of alien organisms thriving during the innovation boom of the Cambrian. These fantastic creatures are far more diverse in their basic plan than the creatures we descended from. What we see since the Burgess Shale, Gould argues, is decreasing diversity of basic plans, with vastly increasing quantities of minor gingerbreading.

For instance, life churns out millions more kinds of insects, in ever more glorious modifications, but no more new kinds of things such as insects. Endless variations of trilobites, but no new classes such as trilobites. And since the Burgess Shale displays a smorgasbord of structural variety that beats the paltry choice of basic plans which life now offers in the same area, one could argue that the conventional view of diversity beginning small and ballooning over time is inverted.

If you count diversity as significant variety, then diversity is shrinking. Some paleontologists are calling this more fundamental diversity of ground plan "disparity" to distinguish it from the ordinary diversity of species. There is more significant difference (fundamental disparity) between a hammer and a saw, than there is between an electric table saw and a power circular saw or all the thousands of baroque electrical appliances manufactured today. Gould puts it this way, "Three blind mice of differing species do not make a diverse fauna, but an elephant, a tree, and an ant do -- even though each assemblage contains just three species." We give more weight to fundamentals of clearly different logic in recognition that it's hard to come up with really innovative basic plans (try to imagine a universal alternative to the tubular gut!).

Because versatile basic plans are rare, when the majority of them go belly up, as they did after the Cambrian, never to be replaced, it's big news. This leads Gould to the "surprising fact of life's history -- marked decrease in disparity followed by an outstanding increase in diversity within the few surviving designs." Take ten designs, throw away nine, and do the tenth one up in a bazillion variations, like beetles. The "cone of increasing diversity" we associate with evolution since the Cambrian, then, is more appropriately figured within the level of species diversity, because more species types are alive today than ever before.

Increasing Numbers of Individuals. There are also more individual organisms in total living now than a billion years ago, or perhaps even a million years ago. Presumably life originated only once, so there was once only the first living organism of Adamlike oneness. Now there are uncounted legions.

There is another important way the sheer number of living entities increases. In a hierarchical manner, supergroups and subgroups create individuals. Bees band together to form a colony, so now the number of individuals total the number of bees plus one superorganism. A person is an individual made up of millions of individual cells which may also be counted and added to the increasing total of individual lives. Each of these cells may have a parasite, thus more individuals. In many overlapping ways, notions of individuals can be nested within each other in the same limited space. So within one cubic volume, a hive of bees with cells and mites and viral infections may have more individuals than the same volume full of bacteria. As Stanley Salthe writes in Evolving Hierarchical Systems, "An indefinite number of unique individuals can exist in a finite material world if they are nested within each other and that world is expanding."

Increasing Specialization. Life starts as a process accomplishing many things in general. Over time a single life is differentiated into many individuals doing more specialized things. Just as a general egg cell differentiates through epigenesis to become a legion of specialized cells, so in evolution animals and plants split up into varieties more dependent on narrower niches. The word "evolution," in fact, originally meant the unrolling development of an egg cell into an embryonic creature. The term was only later applied to organic change over time for the first time by Herbert Spencer, who defined evolution (in 1862) as "a change from an indefinite, incoherent homogeneity, to a definite, coherent heterogeneity; through continuous differentiations and integrations."

The trends listed above can be gathered together with increasing specialization to create the following broad picture: Life begins as one, simple, vague, unformed creativity which, over time becomes more and more fixed into a cloud of precise, inflexible, machinelike structures. Once differentiated, cell lines rarely revert to the more general. Once specialized, animal lines rarely revert to the more general. Over time the percentage of specialized organisms increase, the kinds of specialization increase, and the degree of specialization increases. Evolution moves toward more detail.

Increasing Codependency. Biologists have noticed that primitive organisms have a direct dependency on the physical environment. Some bacteria live inside rock; some lichens eat stone. Slight perturbations of these organisms' physical habitat have a strong impact (lichens are miners' canaries for acid-rain pollution for this reason). As life evolves it unbinds from the inorganic and interacts more with the organic. While plants are rooted directly to the earth, animals, which are rooted to the plants, are freer from the earth. Amphibians and reptiles generally fertilize their eggs and abandon them to the elements, while birds and mammals raise their young, and so are bound closer to life from birth. Over time the close intimacy with earth and minerals is replaced by a dependence on other living things. Parasites cuddling in the warm interior of an animal's gut may never touch anything outside of organic life. Likewise social animals: while ants may live in the ground, their individual lives are far more dependent upon the other ants than upon the soil around them. Deepening sociality is yet another form of life's increasing codependence on other life. Humans are an extreme example of increasing dependence on life rather than the abiotic.

Evolution pulls life away from the inert and binds to itself whenever possible, manufacturing a great something out of nothing.

Increasing evolvability. In 1987, Cambridge zoologist Richard Dawkins presented a paper at the First Artificial Life Workshop entitled "The Evolution of Evolvability," wherein he explored the feasibility and advantages of evolution evolving itself. Around the same time Christopher Wills writing in Wisdom of the Genes, also published a scenario of how genes might control their own evolvability.

Dawkins's thinking was inspired by his attempts to create an artificial evolution in Biomorph Land. He realized while playing God that certain rare innovations would not only provide an immediate advantage to an individual but were "evolutionarily pregnant" and loosened up future offspring's ability to vary widely. He used the example of the first segmented animal in real life which he called "a freak ... [which was] not a dramatically successful individual." But something about animal segmentation was a watershed event that birthed a line of descendants who were champion evolvers.

Dawkins proposed a higher-level natural selection "which favors, not just adaptively successful phenotypes, but a tendency to evolve in certain directions, or even just a tendency to evolve at all." In other words, evolution would select not only for survivability, but also for evolvability.

The ability to evolve does not rest in a single trait or function -- such as mutation rate -- yet a function such as mutation rate will play a role in an organism's evolvability. If a species cannot generate requisite variety, it won't evolve. Its ability to modify its body plays a role in its evolvability, as does its behavioral plasticity. The flexibility of its genome is of critical importance. Ultimately the evolvability of a species is a systems characteristic that does not dwell in any single place, just as an organism's ability to survive does not rest in any single place.

Like all traits selected by evolution, evolvability must be accumulative. A weak innovation once adopted can serve as the platform for the birth of a stronger innovation. In this way, weak evolvability establishes an ongoing base for further evolvability to arise. Over the very long term, evolvability is an essential component of survivability. Thus a line of organisms with genes wired to increase evolvability would accumulate a decided ability (and advantage) to evolve. And so on ad infinitum.

The evolution of evolution is like getting the wish that Aladdin's lamp won't let you have: the wish for three more wishes. It's the power to change the rules of the game legally. Marvin Minsky noticed a similar power of change-which-changes-its-own-rules in the development of a child's mind. Minsky: "A mind cannot really grow very much by only accumulating more and more new knowledge. It must also develop new and better ways to use what it already knows. That's Papert's Principle: Some of the most crucial steps in mental growth are based not simply on acquiring new skills but on acquiring new administrative ways to use what one already knows."

The process by which change is altered is the larger target of evolution. The evolution of evolution does not mean merely that the mutation rate is evolving, although it could entail this. In fact, the mutation rate is remarkably constant over time throughout not only the organic world but also the world of machines and hyperlife. (It is rare for mutation rates to go above a few percent and rare for them to drop below a hundredth of a percent. Somewhere around a tenth of a percent seems to be ideal. That means that a nonsensical wild idea once in a thousand is all that is needed to keep things evolving. Of course one in a thousand is pretty wild for some places.)

Natural selection tends to maintain a mutation rate for maximal evolvability. But for the same advantage, natural selection will move all parameters of a system to the optimal point where further natural selection can take place. However that point of optimal evolvability is a moving target shifted by the very act of reaching for it. In one sense, an evolutionary system is stable because it continually returns itself to the preferred state of optimal evolvability. But because that point is moving -- like a chameleon's colors on a mirror -- the system is perpetually in disequilibrium.

The genius of an evolutionary system is that it is a mechanism for generating perpetual change. Perpetual change does not mean recurrent change, as the kaleidoscope of pedestrian action on a street corner may be said to endure perpetual change. That's really perpetual dynamism. Perpetual change means persistent disequilibrium, the permanent almost-fallen state. It means change that undergoes change itself. The result will be a system that is always on the edge of changing itself out of existence.

Or into existence. The capacity to evolve must be evolved itself. Where else did evolution come from in the first place?

If we accept the theory that life evolved from some kind of nonlife, or protolife, then evolution had to precede life. Natural selection is an abiological consequence; it could very well work on protoliving populations. Once fundamental varieties of evolution were operating, more complex varieties kicked in as the complexity of forms allowed. What we witness in the fossil record of Earthly life is the gradual accumulation of various types of simpler evolutions into the organic whole we now call evolution. Evolution is a conglomeration of many processes which form a society of evolutions. As evolution has evolved over time, evolution itself has increased in diversity and complexity and evolvability. Change changes itself.