The Science of Discworld IV

speak of technology evolving. (The analogy is loose and should not be taken to extremes. Technical drawings or CAD designs are poor analogues of genes.) Selection of technology may appear to be by human agency, but it is highly constrained. Success is decided by vote, and voters vote with their wallets. The inventor’s intentions are almost irrelevant. Just as in biological evolution, the main constraint is
what works
.
    Because of the difficulties inherent in the simple-minded approach to design that Paley proposed – because designers don’t go straight from idea to design to product – we should look carefully at just how designs appear in technology. That changes our attitude to ‘design’ in nature too.
    Most human designs don’t work the first time round. Most jugs still don’t pour well. It’s cheaper to invent a new type of jug, even if it’s no good, than it is to pay licensing fees for one that works. Thebetter mousetrap, even when it is genuinely better, is a minor variation on hundreds of previous mousetraps. Usually.
    Mousetrap evolution is a coherent process, not just a succession of unrelated gadgets. The same goes for bicycles, cars, computers, even jugs. Each new capability causes a particular technological path to branch, leading to new roads. Stuart Kauffman, one of the founders of complexity science, introduced the term ‘the adjacent possible’ to mean the possible behaviours of a complex system that are just a short step away from wherever it currently is. The adjacent possible is a list of what potentially might develop. In a sense, it
is
the system’s potential.
    Organic evolution proceeds by invading the adjacent possible. Invasions that fail aren’t invasions at all, and nothing much changes. Successful invasions don’t just change the system that does the invading; they change the adjacent possible of
everything
. When insects first took to the air, the ones that stayed on the ground were suddenly in danger of predation from above, even though they hadn’t changed at all. Likewise, technology advances by continually invading the adjacent possible. Technological evolution is faster than organic evolution because human minds can use their imaginations to jump into the adjacent possible and see if it works, without actually doing it. They can also copy, which organic evolution does only rarely, aside from reproducing near copies of organisms. These are processes that generate paths and histories, and contexts in which some evolutionary trajectories are viable, but others are not. Only a few select trajectories work. In contrast, thinking in terms of innovations that generate products makes the design process work like magic.
    There are a few useful analogies between technological evolution and organic evolution, and a lot of misleading ones. Comparisons between organic evolution and economics abound in the literature, and nearly all of them are misleading, from social Darwinism to the ‘cost’ of reproduction. Some evolutionary trajectories, however, can usefully be compared to biological ones. Examples include telegraph→ telephone, especially international with undersea cables as investment, pens → word processors, and rockets → space elevators, which we’ll come to shortly. These changes get rid of old constraints at each subsequent recursion.
    There are biological precedents, in which evolution did
not
lead to increased complexity (as measured by DNA information), but the reverse. One is the evolution of mammals. Mammals have
less
DNA than their ‘more primitive’ amphibian ancestors, a trick that can be pulled off because mother mammals control the temperature of their developing embryos by keeping them inside their own bodies. Amphibians need huge quantities of genetic instructions to plan for many different contingencies, as their embryos grow in a pond, subject to the unpredictable vagaries of the weather. Mammals dispense with this excess baggage by investing in temperature control.
    With the expanding possibilities of the chemical/physical universe as a substrate, and organic evolution as a model for an emergent phase space, we should be asking ‘What are the constraints on technology, if any?’ rather than ‘What is the pattern of technological advance?’ Sometimes there are persistent patterns. Moore’s Law states that computing power doubles every eighteen months. It has worked for decades, even though (indeed, because) technologies have changed