Pretty much the most interesting blog on the Internet.— Prof. Steven Landsburg

Once you get past the title, and the subtitle, and the equations, and the foreign quotes, and the computer code, and the various hapax legomena, a solid 50% English content!—The Proprietor

Friday, October 2, 2015

A Partial Explanation for a Genuine Mystery

We need brains

A recent post posited a genuine mystery: Why would evolution design a brain with capacities as useless as, for example, performing higher mathematics? This post attempts to offer a partial answer: The human brain, far from being an efficient and elegant design, is a bit of a bodge job.

Evolution, with its slow mutations, selection for fitness in a particular environment, and sometimes rapid genetic population sweeps, functionally resembles a designer following an algorithm to optimize fitness; a fact from which some have inferred the existence of a designer.

However, it must be noted that such a designer would, unlike most posited deities, not only lack omniscience, but be extremely myopic. For as the designs evolve over the ages, rather than jumping to local or global maxima of fitness, they often stumble around aimlessly in the flatlands of the fitness function or get stuck in local maxima of fitness and never reach a nearby, higher maximum. Apparently, the designer can only see the local gradients of the fitness function, rather than the entire shape.

The most important fact about the evolution of intelligence is that brains (or neurons, until recently the only substrate discovered to potentially support intelligence) are hideously expensive. The human brain, for example, constitutes at most a few percent of body mass and, unlike—let’s say—a muscle, performs no physical work at all. Yet, its support requires fully a third of a human’s metabolism, including respiratory effort and caloric intake!

In environments, like the ones in which almost all of evolution took place, a principal constraint on population growth is Malthusian—there just aren’t enough calories to go around. This renders increases in brain size often evolutionarily unfit. Where the benefits of brains are valuable enough to overcome this cost, brains are still used as sparingly as possible. Over evolutionary time, species improve their fitness by trimming away all brain that is not absolutely necessary for the task. As a consequence, the brain designs of non-human species are sparse and efficient. Even the parts of the human brain shared with other species share in this efficient match to purpose.In that regard, they are more like the restrictive, but efficient, special-purpose logic processors, rather than the unrestricted, but less efficient, general-purpose logic of modern CPUs or the human brain.

A few million years ago, however, our ancestors stumbled on a corner in design space where brains suddenly offered a payoff that justified nearly unlimited investment in it. This corner must not have been a very large part of design space; otherwise evolution would have stumbled into it much sooner. But the fitness gradient for brains must gigantic; for the design evolved rapidly to increase brain size and intelligence until the limiting factor became women’s hip sizes and the necessity of medically unassisted childbirth.

In some ways, evolution reacted to the newly discovered potential payoff for intelligence in the same way a large corporation or government might react to a newly discovered, expensive but enormously rewarding possibility; think of the Manhattan Project. All resources are poured into the effort to be the first to achieve this possibility. Costs are no object. Efficiency be damned. Everything and the kitchen sink are thrown at it.

Such crash projects are sometimes succesful. But a feature universally shared by their products is that while they are workable, they are neither elegant nor efficient. Every conceivable capacity that someone at some stage thought might come in handy is added to the design. Even when ultimately it turns out that some capacities are useless for the task, it is too late to remove them.

So it is, one surmises, with the human brain. All its apparently useless capacities, such as performing higher mathematics, are accidental and undesired byproducts. If human intelligence had evolved on a slower, more deliberate time-scale, such capacities would surely have been optimized away.