Complex evidence, simple decisions
Which way should a pipe go around a building? These decisions are bound by a series of straightforward rules. But when you start to overlay all the elements of a building — structural and decorative, as well as the utilities — routing can start to get complex.
As this post from BuildingSP (shared with me by the ever-alert Matthew Kershaw) points out, detecting clashes between such building components used to be a complex, manual task involving light-boxes and annotated vellum.
The first generation of construction technologies automated the detection of such clashes. But the next generation avoids them altogether, by placing the routing design in the hands of software. Software that can take in the enormous volumes of information about the building and make design decisions within the limited range of options for pipes, cables and mechanical components.
This single example is indicative of some of the natural applications for what might loosely be termed AI technologies, or perhaps more accurately, algorithmic technologies. Applications where the range of evidence (‘big data’, if you like) might overwhelm a human being but where the range of answers is constrained by hard rules.
We’ve only seen the beginning of the application of technologies like this to everyday challenges. Most of the applications to date have been in purely digital situations: categorising images for social networks, personalising web sites or content services. But they have all sorts of applications in the physical world, from packing containers to routing busses.
Wherever there is that great asymmetry between evidence and decision complexity, there is a role for a machine.
*Some people object to this broad and general use of the term AI. But for me if it’s doing white-collar work that would other-wise be done by people, we might as well term it an AI in this context. I understand that the term may need to be used more carefully in other contexts — e.g. research.