Sex may not have succeeded in machine learning, but as a consolation, it has played a prominent role in the evolution of technology in other ways. Pornography was the unacknowledged “killer app” of the World Wide Web, not to mention the printing press, photography, and video before it. The vibrator was the first handheld electrical device, predating the cell phone by a century. Scooters took off in postwar Europe, particularly Italy, because they let young couples get away from their families. Facilitating dating was surely one of the “killer apps” of fire when Homo erectus discovered it a million years ago; and equally surely, a key driver of increasing realism in humanlike robots will be the sexbot industry. Sex just seems to be the end, rather than the means, of technological evolution.
Nurturing nature
Evolutionaries and connectionists have something important in common: they both design learning algorithms inspired by nature. But then they part ways. Evolutionaries focus on learning structure; to them, fine-tuning an evolved structure by optimizing parameters is of secondary importance. In contrast, connectionists prefer to take a simple, hand-coded structure with lots of connections and let weight learning do all the work. This is machine learning’s version of the nature versus nurture controversy, and there are good arguments on both sides.
On the one hand, evolution has produced many amazing things, none more amazing than you. With or without crossover, evolving structure is an essential part of the Master Algorithm. The brain can learn anything, but it can’t evolve a brain. If we thoroughly understood its architecture, we could just implement it in hardware, but we’re very far from that; getting an assist from computer-simulated evolution is a no-brainer. What’s more, we also want to evolve the brains of robots, systems with arbitrary sensors, and super-AIs. There’s no reason to stick with the design of the human brain if there are better ones for those tasks. On the other hand, evolution is excruciatingly slow. The entire life of an organism yields only one piece of information about its genome: its fitness, reflected in the organism’s number of offspring. That’s a colossal waste of information, which neural learning avoids by acquiring the information at the point of use (so to speak). As connectionists like Geoff Hinton like to point out, there’s no advantage to carrying around in the genome information that we can readily acquire from the senses. When a newborn opens his eyes, the visual world comes flooding in; the brain just has to organize it. What does need to be specified in the genome, however, is the architecture of the machine that does the organizing.
As in the nature versus nurture debate, neither side has the whole answer; the key is figuring out how to combine the two. The Master Algorithm is neither genetic programming nor backprop, but it has to include the key elements of both: structure learning and weight learning. In the conventional view, nature does its part first-evolving a brain-and then nurture takes it from there, filling the brain with information. We can easily reproduce this in learning algorithms. First, learn the structure of the network, using (for example) hill climbing to decide which neurons connect to which: try adding each possible new connection to the network, keep the one that most improves performance, and repeat. Then learn the connection weights using backprop, and your brand-new brain is ready to use.
But now there’s an important subtlety, in both natural and artificial evolution. We need to learn weights for every candidate structure along the way, not just the final one, in order to see how well it does in the struggle for life (in the natural case) or on the training data (in the artificial case). The structure we want to select at each step is the one that does best after learning weights, not before. So in reality, nature does not come before nurture; rather, they alternate, with each round of “nurture” learning setting the stage for the next round of “nature” learning and vice versa. Nature evolves for the nurture it gets. The evolutionary growth of the cortex’s associative areas builds on neural learning in the sensory areas, without which it would be useless. Goslings follow their mother around (evolved behavior) but that requires recognizing her (learned ability). If you’re the first thing they see when they hatch, they’ll follow you instead, as Konrad Lorenz memorably showed. The newborn brain already encodes features of the environment but not explicitly; rather, evolution optimized it to extract those features from the expected input. Likewise, in an algorithm that iteratively learns both structure and weights, each new structure is implicitly a function of the weights learned in previous rounds.
Of all the possible genomes, very few correspond to viable organisms. The typical fitness landscape thus consists of vast flatlands with occasional sharp peaks, making evolution very hard. If you start out blindfolded in Kansas, you have no idea which way the Rockies lie, and you’ll wander around for a long time before you bump into their foothills and start climbing. But if you combine evolution with neural learning, something interesting happens. If you’re on flat ground, but not too far from the foothills, neural learning can get you there, and the closer you are to the foothills, the more likely it will. It’s like being able to scan the horizon: it won’t help you in Wichita, but in Denver you’ll see the Rockies in the distance and head that way. Denver now looks a lot fitter than it did when you were blindfolded. The net effect is to widen the fitness peaks, making it possible for you to find your way to them from previously very tough places, like point A in this graph:
In biology, this is called the Baldwin effect, after J. M. Baldwin, who proposed it in 1896. In Baldwinian evolution, behaviors that are first learned later become genetically hardwired. If dog-like mammals can learn to swim, they have a better chance to evolve into seals-as they did-than if they drown. Thus individual learning can influence evolution without recourse to Lamarckism. Geoff Hinton and Steven Nowlan demonstrated the Baldwin effect in machine learning by using genetic algorithms to evolve neural network structure and observing that fitness increased over time only when individual learning was allowed.
He who learns fastest wins
Evolution searches for good structures, and neural learning fills them in: this combination is the easiest of the steps we’ll take toward the Master Algorithm. This may come as a surprise to anyone familiar with the never-ending twists and turns of the nature versus nurture controversy, 2,500 years old and still going strong. Seeing life through the eyes of a computer clarifies a lot of things, however. “Nature” for a computer is the program it runs, and “nurture” is the data it gets. The question of which one is more important is clearly absurd; there’s no output without both program and data, and it’s not like the output is, say, 60 percent caused by the program and 40 percent by the data. That’s the kind of linear thinking that a familiarity with machine learning immunizes you against.
On the other hand, you may be wondering why we’re not done at this point. Surely if we’ve combined nature’s two master algorithms, evolution and the brain, that’s all we could ask for. Unfortunately, what we have so far is only a very crude cartoon of how nature learns, good enough for a lot of applications but still a pale shadow of the real thing. For example, the development of the embryo is a crucial part of life, but there’s no analog of it in machine learning: the “organism” is a very straightforward function of the genome, and we may be missing something important there. But another reason is that we wouldn’t be satisfied even if we had completely figured out how nature learns. For one thing, it’s too slow. Evolution takes billions of years to learn, and the brain takes a lifetime. Culture is better: I can distill a lifetime of learning into a book, and you can read it in a few hours. But learning algorithms should be able to learn in minutes or seconds. He who learns fastest wins, whether it’s the Baldwin effect speeding up evolution, verbal communication speeding up human learning, or computers discovering patterns at the speed of light. Machine learning is the latest chapter in the arms race of life on Earth, and swifter hardware is only half the equation. The other half is smarter software.