Is bioengineering the best route to nanotechnology? The words "nanotechnology" or "molecular manufacturing" conjures up many different sorts of images to different people. But I think the two ideas of nanotech that excite the greatest number of people are: "little microscopic robots that self-replicate and then roam around doing stuff" (that is, bacteria); "factories that make useful products out of sunlight, earth, and water, and can also make other factories of the same kind" (that is, plants). It seems to me that genetically modifying natural bacteria and plants so that they better serve our purposes is going to be orders and orders and orders of magnitude easier than building robot bacteria and robot plants, if you will, from scratch. A possible analogy: Two cavemen, tired of lugging stuff around, are trying to think up better ways to do things. Alf says, "I'm gonna build a mechanized rolling machine. I'll put all my stuff in the luggage compartment and drive around real fast. Just need to figure out how to make the engine." Bob says, "Animals are sort of like natural engines. I'm gonna capture some wild animals--dogs, asses, oxen, horses maybe. I'll modify them by breeding to make them bigger, or faster, or more docile. Then I'll modify them some more by connecting them to apparatus like sledges, carriages, plows, millstones, and get them to do stuff for me." Alf says, "Don't screw around with those evolved systems. You don't even understand how a horse works. It's better to design everything yourself from scratch so you have some intuitive understanding of it. You'll see, my horseless carriage is the way to go." Ultimately Alf was right and his mechanized carriage caught on, but he did not live to see it. Meanwhile, Bob's ideas proved to be of more practical use for a long time-- several millenia. I also think that natural life forms should serve as a good reality check for what nanotech robots will be able to do. If you think that your microscopic robots will be faster and stronger and just generally much cooler than bacteria are, or if you think that your solar-powered self-erecting factory will rise out of the ground much faster than, say, a bamboo plant does, then I think you are suffering from some hubris and will probably be disappointed. We can build skyscrapers taller than redwoods, but this is not so much because we are smart and evolution is dumb, but because in building skyscrapers we cheat in numerous ways, notably by making them out of steel. Redwoods build themselves out of only local raw materials. Skyscrapers require iron to be dug out of the ground (probably not locally), melted, and made into steel with considerable energy expenditure, before construction can even begin. Redwoods are their own construction equipment and bootstrap themselves up from a seed using only solar power. Skyscrapers require pre-existing cranes, trucks, earthmovers, etc., none of which limit themselves to solar power. (Interestingly, even after all that cheating, the tallest skyscrapers in the world are only about four times taller than a typical redwood. Not 100 times taller, not even ten times taller. So for a caveman speculating about the height of skyscrapers, a redwood really wouldn't be such a terrible guide. For what it's worth, the largest wooden building in the world is only one-half as tall as a typical redwood.) We can build airplanes which are much bigger and faster than any living bird. But in the size and speed regimes in which they do operate, birds are infinitely more sophisticated than airplanes. A bird-sized model airplane has nowhere near the maneuverability of a bird. Imagine an airplane flitting from tree branch to tree branch. Imagine an airplane flying through a hedge. How far would an airplane get if it had to scavenge its own jet fuel from the wild? In general, up to this point in history anyway, when our machines seem to outperform comparable products of evolution, it is usually not because we have found some better design that evolution overlooked. Plants and animals are astoundingly sophisticated and well-designed. If our machines are better, it is usually because we can change the rules of the game. We can "cheat" by giving our machines materials and energy sources not naturally available in their operating environment, and/or by giving them unnaturally friendly environments to operate in (e.g. roads for cars). Also, the machine's goal may be different from the animal's goal: perhaps for the airplane, nothing matters except maximum speed, whereas the bird has to satisfy many more different performance criteria. But if you imagine building a nanotechnological robot with the mission, "Go out into the world and build copies of yourself using the raw materials and resources available in your environment," you are describing very much the job for which biology has been optimizing itself. You have to satisfy much the same performance criteria as organic life forms and do it with the same resources. Biology's laboratory space is the size of the world, its laboratory budget is the amount of solar energy falling on the Earth, and its research program in this area has been ongoing for hundreds of millions of years. Therefore you should not casually assume that your research group has the brains to do a better optimization job. You might say, why not "cheat" on behalf of our nanotechnological machines too? E.g., let's not require our self-assembling factory to grow in natural dirt, we will give it a prepared pit full of pure graphite to build in. Let's not require it to photosynthesize, instead we will provide it with bottles of kerosene to drink. Or whatever. I'm sure we could and will do those sort of things, but, to my mind anyway, that is already a retreat from the most grandiose popular ideas about self-sufficient nanobots.

Richard Mason

---