Incongruous technologies redux

Good news: I have worked out why my incongruous technologies are incongruous: it's because of software!

I just watched an excellent documentary on mechanical automata. It includes demonstrations (among several other excellent examples) of "The draughtsman" and "The writer" built by Pierre Jaquet-Droz in the late 1700s. These automata are little mechanical boys that can draw pictures and write text, respectively.

I wrote in the previous post on this topic that we don't have any applications, beyond watchmaking, for systems of tiny springs and gears. But these automata are excellent examples to the contrary. Just look at the back of "The writer":

In particular, note the large circle at the bottom: this is how the device is programmed. Each of the shapes loaded around the perimeter of this circle has a different height, and the height changes how far the stack of discs in the middle of the boy is raised up, and the height that this stack is raised up decides the set of discs that are selected, and the set of discs that are selected decides the letter that the writer will draw. The whole stack of discs rotates, each of the boy's "muscles" follows its selected disc, and the boy writes the text that you've programmed in to him. Superb.

This is exactly the kind of stuff I thought we didn't have, and they were making it 250 years ago!

So why don't we make stuff like this today?

Well... imagine how you would go about designing "The writer" if you were designing him today. Unless you're crazy, or have a very specific and unusual set of skills, or want to do it for historical interest instead of practical purposes, you wouldn't do it like this. You'd have a dinky little microcontroller and 3 stepper motors and you'd save yourself a load of time and money and effort and frustration.

The reason people used to make stuff like this isn't that they had tapped into remarkable technology that we're under-utilising now. It's that they didn't have software! If Pierre Jaquet-Droz could have solved his problems using software he would have done so, but he didn't have software so he had to make up for it with incredible ingenuity and hard work.

So the real question isn't "why don't we make better use of tiny gears and springs?" It's "why do we still use gears and springs at all?".

And this also gives us an answer as to why we don't make better use of IC fabs: there's no reason to make anything other than integrated circuits, because integrated circuits are just a substrate for software, and software gets you everything you want.

So why do we still use gears and springs at all? Why don't we do literally everything with software? At some point (for the time being!) we still need to interact with the physical world. For our digital clone of "The writer" boy to actually write on actual paper, he needs to actually move an actual pen, and to do that he'll need motors and probably gearboxes.

But the kinds of springs and gears that are involved there are substantially simplified compared to what Pierre Jaquet-Droz was working with. Instead of making bespoke gears and springs for every application, we make generic gears and springs, and control them with bespoke software, because software is quick and easy to make and mechanical systems aren't.

Obviously, in the long-term, we're all going to be living our lives in permanent VR (to the extent that experiencing the world through the filters of our senses and brain is not already a kind of permanent VR). And then we won't be needing any more motors.

(Also, note that the kind of "pen" that Jaquet-Droz uses is not what we would recognise today, because actual pens hadn't been invented back then. So "The writer" has to automatically dip his quill in the ink periodically while he does his writing! And this is all orchestrated with gears and springs and cams and shafts).

Anyway, the short answer as to why we aren't making maximal use of tiny mechanical systems, or of IC fab technology, is because we can make use of software to solve the same problems faster, cheaper, and easier. It might be interesting to see how that can be extended further: next time you encounter a tiny bespoke mechanical system, maybe imagine how it could be replaced with a generic mechanical system controlled by software.