A new essay by Steve Jurvetson in MIT'sTechnology Review asks, "How would we build a really complex system -- such as a general artificial intelligence (AI) that exceeded human intelligence?"
He then suggests three possibilities: "Some technologists advocate design; others prefer evolutionary search algorithms. Still others would conflate the two, hoping to incorporate the best of both while avoiding their limitations."
By evolutionary search algorithms, he means a process like Darwinian evolution, in which natural selection seizes beneficial random variations and passes them along to future generations in such a way that these beneficial variations gradually accumulate into ever more sophisticated designs.
Jurvetson notes that whereas human designed systems "tend to break easily, and they have conquered only simple problems so far," the typical biological system is, by comparison, "robust, resilient, and well adapted to its environment."
The choice then sounds like a no-brainer: use evolutionary search algorithms: they yield better designs. There's just one problem, and it's the elephant in the room of modern evolutionary theory. Jurvetson deserves credit for pointing it out in such plain spoken language:
In fact, biological evolution provides the only "existence proof" that an algorithm can produce complexity transcending that of its antecedents.
Why is this an elephant rather than a mouse?
Because no one has ever witnessed biological evolution actually doing this, which means there are no existence proofs "that an algorithm can produce complexity transcending that of its antecedents." That's not just a problem for Darwinism but for all origins models rooted in the belief that organized and functional complexity can evolve from simplicity without intelligent guidance.
Even instances of microevolution in which bacteria develop antibiotic resistance do not involve net increases in biological complexity. As University of Idaho biologist Scott Minnich has demonstrated in repeatable laboratory work, the resistant bacteria sacrifice more general fitness on the way to becoming immune to the antibiotic. Place the resistant bacteria back in a pool of the ordinary bacteria, where the threat of the antibiotic has been removed, and the "more primitive" bacteria quickly overwhelm the specialists.
Human designers surely will continue to pick up design ideas from the many ingenious designs found and still to be found in the biological realm. Scientists may even master nature's ability to harness minor random variations to allow self-reproducing systems to adjust in limited but significant ways to environmental changes. In the burgeoning field of biomimetics, humans are very much the apprentices. Pretending the master is an algorithm will do nothing to speed, and much to slow, the learning curve.
