Cheaper than printing it out: buy the paperback book.

Out of Control

Only in the last couple of years has the exhilarating link between learning, behavior, adaptation, and evolution even begun to be investigated. Most of this exciting work has been performed in computer simulations. It has been more or less ignored by biologists -- which is not the stigma it once was. A number of researchers such as David Ackley and Michael Littman (in 1990), and Geoffrey Hinton and Steven Nowlan (in 1987) have shown clearly and unequivocally how a population of organisms that are learning -- that is, exploring their fitness possibilities by changing behavior -- evolve faster than a population that are not learning. In the words of Ackley and Littman, "We found that learning and evolution together were more successful than either alone in producing adaptive populations that survived to the end of our simulation." Their organism's exploratory learning is essentially a random search of a fixed problem. But in December 1991, two researchers, Parisi and Nolfi, presented results at the First European Conference on Artificial Life which showed that self-guided learning -- where the problem task is selected by the population themselves -- produced optimal rates of learning, which in turn may increase adaptation. They make a bold claim, which will be heard more and more in biology, that behavior and learning are among the causes of genetic evolution.

There is a further caveat. Hilton and Nolan surmise that Baldwinism most likely works only on severely "rugged" problems. They say, "For biologists who believe that evolutionary spaces contain nice hills...the Baldwin effect is of little interest, but for biologists who are suspicious of the assertion that the natural search spaces are so nicely structured, the Baldwin effect is an important mechanism that allows adaptive processes within the organism to greatly improve the space in which it evolves." The organism creates its own possibilities.

"The problem with Darwinian evolution," Michael Littman told me, "is that it is great if you have evolutionary time!" But who can wait a million years? In the collective effort to introduce artificial evolution into manufactured systems, one way to accelerate the speed at which things evolve is to add learning to the soup. Artificial evolution will probably require a certain amount of artificial learning and intelligence to make it happen within human time scales.

Learning plus evolution is basically the recipe for culture. It may be that just as learning and behavior can pass off their information to genes, genes can pass their information off onto learning and behavior. The former is called genetic assimilation; the latter, cultural assimilation.

Human history is a story of cultural takeover. As societies develop, their collective skill of learning and teaching steadily expropriates similar memory and skills transmitted by human biology.

In this view -- which is a rather old idea -- each step of cultural learning won by early humankind (fire, hammer, writing) prepared a "possibility space" that allowed human minds and bodies to shift so that some of what it once did biologically would afterwards be done culturally. Over time the biology of humans became dependent on the culture of humans, and more supportive of further culturalization, since culture assumed some of biology's work. Every additional week a child was reared by culture (grandparent's wisdom) instead of by animal instinct gave human biology another chance to irrevocably transfer that duty to further cultural rearing.

Cultural anthropologist Clifford Geertz sums up this hand-off:

"The slow, steady, almost glacial growth of culture through the Ice Age altered the balance of selection pressures for the evolving Homo in such a way as to play a major directive role in his evolution. The perfection of tools, the adoption of organized hunting and gathering practices, the beginnings of true family organization, the discovery of fire, and most critically, though it is as yet extremely difficult to trace it out in any detail, the increasing reliance upon systems of significant symbols (language, art, myth, ritual) for orientation, communication, and self-control all created for man a new environment to which he was then obliged to adapt....We were obliged to abandon the regularity and precision of detailed genetic control over our conduct..."

But if we consider culture as its own self-organizing system -- a system with its own agenda and pressure to survive -- then the history of humans gets even more interesting. As Richard Dawkins has shown, systems of self-replicating ideas or memes can quickly accumulate their own agenda and behaviors. I assign no higher motive to a cultural entity than the primitive drive to reproduce itself and modify its environment to aid its spread. One way the self-organizing system of culture can survive is by consuming human biological resources. And human bodies often have legitimate motivation in surrendering certain jobs. Books relieve the human mind of long-term storage rents, freeing it up for other things, while language compresses awkward hand-waving communication into a thrifty, energy conserving voice. Over generations of society, culture would assimilate more of the functions and information of organic tissue. Sociobiologists E. O. Wilson and Charles Lumsden used mathematical models to arrive at what they call the "thousand-year rule." They calculated that cultural evolution can pull along significant genetic change so that it catches up in only a thousand years. They speculate that the vast changes we have seen in our culture over the last millennium could have some foundation in genetic change, even though genetic change might not be visible.

So tightly coupled are genes and culture, Wilson and Lumsden say, that "genes and culture are inseverably linked. Changes in one inevitably force changes in the other." Cultural evolution can shape genomes, but it can also be said that genes must shape culture. Wilson believes that genetic change is a prerequisite for cultural change. Unless the genes are flexible enough to assimilate cultural change, he believes it will not take root for the long term.

Culture follows our bodies, while our bodies follow culture. In the absence of culture, humans seem to lose distinctly human talents. (As somewhat unsatisfactory evidence we have the failures of "wolf children" raised by animals to develop into creative adults.) Culture and flesh, then, meld into a symbiotic relationship. In Danny Hillis's terminology, civilized humans are "the world's most successful symbionts" -- culture and biology behaving as mutually beneficial parasites for each other -- the coolest example of coevolution we have. And as in all cases of coevolution, it implies positive feedback and the law of increasing returns.

Cultural learning rewires biology (to be precise, it allows biology to remodel itself) so that biology becomes susceptible to further culturalization. Thus, culture tends to accelerate itself. In the same way that life begets more life and more kinds of life, culture begets more culture and more kinds of culture. I mean it in a strong way, that culture produces organisms that are biologically more able to produce, learn, adapt in cultural ways, rather than biological ways. This implies that the reason we have brains that can produce culture is that culture produced brains that could. That is, whatever shred of culture resident in prehuman species was instrumental in molding offspring to produce more culture.

To the human body this accelerating evolution towards an information-based system looks like biological atrophy. From the view of books and learning, it looks like self-organization, culture amplifying itself at the expense of biology. Just as life infiltrates matter mercilessly and then hijacks it forever, cultural life hijacks biology. In the strong sense I'm advocating here, culture modifies our genes.

I have absolutely no biological evidence for all this. I've heard casual things from folks like Steven Jay Gould who says the "morphology of humans hasn't changed in the 25,000 years from Cro-Magnon," but I don't know what that means for this idea, and how true his assertion is. On the other hand, devolution is weirdly quick. Lizards and mice can lose their eyesight in a blink (so to speak) inhabiting lightless caves. Flesh, it seems to me, is ever ready to give up part of its daily grind if given a chance.

My larger point is that the advantages of Lamarckian evolution are so great that nature has found ways to make it happen. In Darwin's metaphor I would put its success this way: Evolution daily scrutinizes the world not just to find fitter organisms, but to find ways to increase its own ability. It hourly seeks to gain an edge in adaptation. Its own ceaseless pushing creates an immense pressure -- like the weight of an ocean seeking a crack to seep through -- to increase its adaptive abilities. Evolution searches the surface of the planet to find ways to speed itself up, to make itself more nimble, more evolvable -- not because it is anthropomorphic, but because the speeding up of adaptation is the runaway circuit it rides on. It searches for the advantages of Lamarckian evolution without realizing it because Lamarckism is a crack of less resistance and more evolvability.

When animals with complex behavior evolved, evolution began to break out of its Darwinian straight jacket. Animals could react, choose, migrate, adapt, and give room for the blossoming of pseudo-Lamarckian evolution. As human brains evolved, they created culture, which permitted the birth of a true Lamarckian system of inherited acquisitions.

Darwinian evolution is not just slow learning. In Marvin Minsky's words, "Darwinian evolution is dumb learning." What evolution later found in primitive brains is a way to quicken itself by introducing learning into the equation. What evolution eventually found in the human brain was the complexity needed to peer ahead in anticipation and direct evolution's course.