Читать книгу Standing on the Sun - Christopher Meyer - Страница 23

We've already described (in chapter 1) the different world in which capitalism now finds itself. It's a world of values, needs, and technologies that certainly have not been prominent in the past. So striking is the difference that capitalism's critics, and some of its adherents, fear it will never be equal to the challenge. These are the people who, whether they would put it this way or not, see capitalism as a machine. They imagine something like a car, designed for paved roads, now having to perform in an ocean. Unable to retool itself, it will engage its gears, spin its wheels, and sink.

A Buick in the Pacific exhibits a disastrous lack of fitness for its milieu. But an AquaBuick's inventor might notice that internal combustion engines work in boats, and tires could serve as life preservers, and find another way to make a living. And in nature, when species find themselves in changing environments, some of them indeed do adapt.

The ones that have the best chance of responding to change are those that have a lot of variations in their gene pools to start with. Individuals that had quirky, even hampering, features in the old world find themselves relying on those features to survive. And if they live long enough to produce offspring, they pass down those newly useful features.

The perhaps depressing fact is that this is the whole purpose of sex: to ensure that there are always enough innovative designs with the potential either to improve on the previous solutions or to suit a newly changed environment better than the old model. Nature arranges to take two sets of instructions for making an organism from the current generation—the parents' DNA—and recombine them in a structured but somewhat random way to make a new set of instructions.⁶ Then the new genotype grows the phenotype, as a new blueprint gives rise to a building.

Next, of course, the environment, through selection, decides whether this new variant was a good idea. The phenotypes that work well in the environment are fit, and the fittest survive and get a preferential opportunity to breed the next generation. The elements of its genotype get a chance to “radiate,” or show up in multiple successor species. (It's the same basic phenomenon when a product thrives that has a certain characteristic, and then that characteristic gains prevalence. This is why, post iPhone, touch screens are becoming ubiquitous.) Evolution is the outcome of new ideas presented to the environment by recombining old ideas and waiting for selection to operate.

If the environment is unchanging, the same genes will prove fit again and again. If something suddenly changes, though, the chance of survival is increased if there's a reservoir of alternatives to fall back on. This is why the U.S. Department of Agriculture, noting that farmers favor only about half a dozen strains of very productive corn, maintains a pool of other corn varieties; it is hedging against the possibility of a new blight, climate change, a new fertilizer or pesticide, or some other shift in the environment. Diversity provides insurance against change, as any financial advisor will tell you. When nature is left to its own devices, sexual recombination preserves a level of diversity in any species' gene pool.

If genetic variation is the most important trait that improves the odds of successful adaptation, then second most important is speed: the more often the available variety of genes is recombined, the faster change can occur. There's a new flu virus every year because influenza can respond to last year's vaccine by changing its shape—so flu survives. Dinosaurs' long gestation periods reduced their ability to survive a sudden climate change.

More depression: sexual recombination is a process that can be precisely described mathematically and is therefore not at all limited to biology. In the 1990s people like Melanie Mitchell (then at the University of Michigan) demonstrated this by taking lines of computer code and “breeding” them to produce new versions that were then tested for fitness for a task—essentially treating binary code as if it were DNA code. DNA is a code with four letters; digital code has only two. But it's not hard to imagine starting with two equally long strings of ones and zeros and creating different ways of taking half the code from each to make a new string. Then that string's fitness to do its intended task can be evaluated and it can have the opportunity, or not, to breed a next generation. The Wikipedia entry on genetic algorithms, as such processes came to be labeled, now lists sixty-two realms, from “artificial creativity” to “wireless sensor/ad-hoc networks” in which they have been applied.⁷

Let's suppose the digital strings a genetic algorithm started with comprised the blueprint of an online ad, for example. And let's agree to define the fitness of an ad as its success at attracting clicks. The Web users who see it are its environment; they do the selecting.⁸ The marketer doing the advertising might produce different versions featuring various combinations of images, copy, and promotional offers. Then by sending these different versions to samples of its market, it would discover which versions were most fit. So far, a pretty standard procedure: this is how marketing is done, whether it's testing alternative magazine covers or tweaking product descriptions in catalogs.

But now imagine that, like a horse breeder, that marketer took the highest-performing ads and applied a genetic algorithm to produce new lines of code for new ads, recombining the different elements (images, copy, offer) that were part of the mix in the first-round winners. Those new genotypes would yield new phenotypes: different-looking ads that might or might not outperform their parents. Now it's time for a round 3 competition—a trivially easy exercise in an online setting—after which the winners produce another batch of progeny. In no time, the marketer has the Triple Crown winner of ads.

This is not a product of our febrile imaginations. We've just described part of the business of Affinnova, an award-winning Massachusetts company that lists Procter & Gamble, Walmart, Diageo, Microsoft, and many other consumer marketers among its clients. Affinnova's strength is in translating recombination and selection to mathematical formulas applying them to design tasks, and executing these inexpensively on the Web. As the company describes it, “Affinnova's technology mimics the process of evolution, but at a greatly accelerated rate.”⁹

Acceleration is the key, given the number of permutations that end up being produced. In one assignment, for example, office-supply company Staples wanted to create packaging for a line of recycled paper products that would appeal to customers. It was flexible on the choice of colors and fonts, and also had certain words it could alter (such as changing multiuse to multipurpose). Affinnova proceeded to serve up design options, in each round asking 750 consumers to choose among three packaging designs. As winners begat more options, increasingly fit designs emerged. In short order, Affinnova analyzed some twenty-two thousand different designs and declared one most fit.

In an impressive proof-of-concept project, the same approach allowed the company to predict political preferences. Applying its “evolutionary optimization technology” to arrive at a winning strategy for the 2008 U.S. presidential election, it asked thousands of likely voters months before the election to react to different combinations of platform issues and vice presidential candidates. The project revealed that the economy would be the key issue with voters, and the Iraq war would not.

The genetic algorithm approach has been applied to dozens of business problems by numerous companies. In one of the most surprising cases, GE used it to design the engine it was building for the Boeing 777. Engineers were startled by the “creativity” shown. Beyond optimizing the parameters of the engine in its standard seven-stage compressor configuration, the software made a leap to a six-stage design, leading to a material improvement in the weight of the engine.

Our main reason for describing these tools, however, is that they vividly illustrate how better solutions can emerge when complex systems are broken into their defining parts, the parts are recombined and subjected to fitness testing, and the process is iterated many times. If you translate this to the notion of capitalism adapting, the defining parts are its current rules and technologies, in all their rich variety. The combinations are tested every day, in diverse environments. NeuroSky CEO Stanley Yang tells us that the empowering, consensus-seeking approach he must employ in Japan would be regarded as wimpy by his team in China and would lose him his employees' respect. What is considered crony capitalism in Israel is legalized as lobbying in the United States. As the landscape of commerce changes, some of these local variants will prove to be valuable genotypes—differences that confer a fitness advantage. Some elements that formerly seemed secondary, or even hampering, may take on central importance.

We should quickly add, too, that we don't expect any single form of capitalism to prevail globally. The mutations will continue, and variation will persist. We return to this theme in chapter 8 as we address the expectation of convergence we've encountered in some quarters. Yes, it's true in genomes as well as social systems that some really good ideas—say, the way the cells of all vertebrates convert food to chemical energy—can become universal. But it's not a contradiction to say that some ideas will prove fit only in some locales. If you took Anthro 101 you likely learned about the sickle cell, which brings with it improved resistance to malaria and therefore is selected for in geographies where malaria is common. Thus it became a common component of the human genome in Africa's lowlands. In another environment or in another time, it serves no good purpose; quite the opposite, sickle cell anemia is a terrible disease. Hence over time the gene becomes less prevalent in the pool.

By analogy, there can be rules that confer advantages in some business environments and not others. Risk aversion, for example, might be an attribute associated with high performance in the heavy industries of the Industrial Revolution. In a steel or chemical plant, a slipup can kill people.¹⁰ In another industry more dependent on constant innovation and on manipulating only bits, risk-aversion DNA could be hobbling. We can't help noting what seems to be a persistent cultural variation in attitudes toward risk in the U.S. auto industry versus the software industry, and it makes sense: the consequences of “crashes” in the two industries are quite different.

And consider what happens when these risk attitudes collide, as in software for medical devices. New attitudes breed at the intersections. Such recombinations are an example of the yeasty, distributed process by which the overall system changes, one rule at a time.

It's an important and fundamental point that the rules of capitalism are selected, not derived from first principles. Whether intentionally, by the World Trade Organization, the U.S. Congress, or international financial regulators, or unwittingly, by the choices made by customers, investors, and executives, the rules continually change and coevolve. A case of fraud induces enforcement of the Sarbanes-Oxley Act, which in turn, as a matter of self-protection, may drive greater collaboration in the accounting profession. And when capitalism finds itself in a new environment, the choices change, because they are made by people having different criteria, living in different circumstances. What is privacy? How should digital property be protected? What corporate behavior should be regulated, and what constitutes monopoly power? What are the rights of different stakeholders in society? There are many possibilities for the answers, just as there are many layouts for an ad.

Perhaps we prefer to think that we decide these questions by rational political debate or judicial processes. But these are only two of the tests of fitness among the interlocking feedback loops of coevolution. And more often than not, they follow practice rather than lead it.