Look at it this way: if you really understand how biological life works, you can use that understanding to do genetic engineering. Here’s an aphorism I learned from studying Philosophy of Science. Explanation equals Prediction equals Technology. That is, Explanation of what has happened, equals Prediction of what will happen, equals Technology of how to make it happen. If you cannot Predict and do Technology, then the Explanation was false in the first place.

At the time of the Second World War, all kinds of very highbrow types, such as pure mathematicians, and physicists, philosophers, psychologists, artists, (classical) musicians, and even electrical engineers, got into the war effort, to stop Hitler and all that. It wasn’t just the hard scientists, of course. Anthropologists were apt to get parachuted behind enemy lines to organize the natives as guerrillas. At any rate, the scientists and suchlike were often put through crash courses in engineering and set to work, mostly on the more physics-oriented types of weapons. There was the Manhattan project, the atomic bomb, of course, but a lot of them wound up working on things like radar and sonar, or code-breaking. At the same time, this exposure to engineering got them thinking in their spare time about whether they could take an engineer’s approach to the ideas they had been thinking about back in their universities. This was the time and place when the electronic computer emerged. In particular, there was the philosopher Noam Chomsky. He began thinking about how one could explain human language and thought in “tinkertoy terms.” There had been a body of linguists before the war (the Prague School) who had been trying to formulate general laws of language, but Chomsky took it a step further, inventing Generative Grammar, the beginning of Artificial Intelligence. In the environment of MIT’s wartime laboratories, people shared ideas back and forth, and the idea of Generative Grammar cross-pollinated into the mind of a mathematician named John Backus, who developed the idea into the FORTRAN compiler. Chomsky’s students, and disciples, and anti-disciples, unto the third and fourth generation, pursued these merged ideas. Eventually, the wartime cohesion began to wear thin, especially under the strain of the Vietnam War. Chomsky eventually got into a titanic quarrel with Marvin Minsky, who was MIT’s major recipient of military DARPA funds for computer research. Probably the single biggest factor separating academic research from business practicality was cost. People were playing around with systems which, if fully implemented, would have required computers costing a trillion dollars, or some similarly modest sum. As costs came down over twenty or thirty years, these systems became much more practical than their original authors had ever imagined. Thus things like Google became possible.

Look at the “qsort” function in the C programming language– that is a fairly good example of what Computer Science is all about. You pass “qsort” some parameters, viz, the starting location of an array, the size of the records in the array, the number of records, and the address of a function– let’s call it “foobar”– which can tell “qsort” which of two records should go first. “qsort” calls “foobar” over and over again until the array is completely sorted. Sorting is treated as a kind of summation of the act of comparison. “qsort” does not know anything about the internal meaning of the records. If you are writing an implementation of “qsort,” you don’t need to know how many records there will be, or how long each record will be. You arrange to automatically copy that information from the input parameters. What you do know about the world outside the “qsort” routine is the precision of the standard integer– whether it is sixteen-bit, or thirty-two-bit, or sixty-four-bit. You also need to know the conventional method of representing machine addresses, which is not always the same as the standard integer. That is about all you need to know.

What I did was to attempt to address a philosopher’s problem by designing a type of “universal inference engine,” in the tradition of Chomsky, at about the same level of abstraction as “qsort.” At the time, the cost of actually building the thing would have been prohibitive, but as costs came down, with Moore’s Law, it became inconvenient prior art for anyone who might want to patent the idea of a universal inference engine. I don’t think I was particularly unique– lots of other young men were doing the same kind of thing at the same time, at various colleges and universities across the country. However, a more or less chance event resulted in my getting a paper publication out of it, and eventually meeting the legal definition of prior art. Bachelors’ theses and similar work were not normally published at the time, this being long before the internet.

If you were designing transmissions at that level, you would be creating a kind of archetype-transmission which would be equally suitable for an automobile, or a motor scooter, or a tractor, or a railroad locomotive. Possibly, you could create a CAD/CAM program which would design transmissions to order from their performance specifications. There are programs which design electronic circuit boards on that principle. If the automakers were to start using electric drive, the way it is used in certain kinds of large equipment (every wheel has its own motor, getting electricity from a common power bus fed by an onboard electric generating plant), you might find yourself using semi-surplus automobile parts for the usual excellent reasons (*), and the distinctive characteristics of tractor transmission design might be reduced to control systems.

(*) A typical production run for an automobile working part is likely to be five or ten million units, and the price accordingly low, and you need compelling reasons not to use such parts which, from the standpoint of specialist equipment, are effectively free and disposable. In electronics, over the years, a lot of specialist companies reached the point of saying that, given the development of small computers, there was no longer any good reason for them to be in the hardware business. They could not make as good a microprocessor for the money as Intel could, so they switched to software instead. It would be interesting to see if that applies to other kinds of industries as well. You probably know about “genset” railroad locomotives using multiple truck engines. There are about 20,000 railroad locomotives, versus about two million eighteen-wheeler trucks. That could be a problem for the specialist firms which make 3000-4000 hp diesel engines.

Well, I understand that sizable electric motors are already “substantially efficient,” ie 90%+ at greater than 25 HP. In short, this puts them in the same general range as gear pairs. Of course, most of the competitive motor market is for much smaller motors, the kind that drive refrigerators, air conditioners, etc. A lot of the discussion of large electric motor efficiency tends not to involve motors per se, but motors in particular applications, and it tends to boil down to not doing the equivalent of running a motor against a brake, that is, fitting them with adjustable speed controllers instead.

The issues of size, weight, and cost are something else, of course.

The equations pertaining to an electric motor ultimately depend on the voltage you choose. If you look at two “best-of-class” projects, the Prius and the GM EV1, you find that they both run at 200-300 volts, the EV1 being nominally rated at 320/220 volts, and the Prius at 273 volts. At one point, it was proposed that the EV1 should be at least partially stepped up to 440 volts. Of course, increasingly stringent safety mechanisms are required as the voltage goes up, but those do not seem to present fundamental difficulties. Light rail (trolleys, subways, etc) operates in part at 660 volts, with reasonable safety, and of course, regular electric railroads run, in part, at much higher voltages. The power is stepped down to 220 volts to be fed through a flexible connection to the wheel motors, of course. When you ride a heavy commuter rail car on the Northeast corridor, the kind that has a pantograph (eg. Philadelphia’s SEPTA system), you may very well be standing within five feet or so of 25,000 volts. The “‘copper economy” of the motors of the two automobiles cited above was sufficient that it didn’t seem to be an identifiable limiting factor. The limiting factor of the EV1 was of course its batteries, the same as for every other pure electric car.

Now, as for cost, that is, generally speaking, going to be a matter of labor inputs. Simply producing things in very large quantities means that all kinds of specialized automation becomes feasible to eliminate hand labor, and the R&D, tooling, and overhead expenses can be spread out over a very large number of units. In effect, low cost is something of a self-fulfilling prophecy.

See:
Michael Shnayerson, The Car That could: The Inside Story of the GM’s Revolutionary Electric Car, 1996