Drones, Silicon Valley and biology: The future isn't here yet

There is an illuminating article in the WSJ that lays out the problems with routine drone delivery that have been plaguing companies like Amazon and Google. Turns out it's one thing to make drones fly, quite another to make them deliver well defined objects in even better defined locations.

Most of the problems with drone delivery that the article highlighted are not too surprising when you think about them. The drones have problems landing smoothly, their GPS has problems pinpointing the precise locations of homes and distinguishing obstacles from landing spots, and they can get caught in or destroyed by any number of obstacles, from power cables to flying birds. There are also some interesting social problems involved: for instance the engineers have to worry about whether people might be scared by drones or, conversely, be too enamored of them and try to steal them. The bottom line is that landing drones on a routine basis in heavily populated residential areas is a messy and unpredictable process that has turned out to be far more challenging than what it seemed to be.

It seems to me that the problems with landing drones could serve as a metaphor for Silicon Valley attempting all kinds of things beyond its core areas of expertise, most notably biology. Just like residential areas, the interiors of cells are crowded, messy and wet environments with water molecules, proteins and small molecules sloshing around against each other. Just like the drone GPS has a problem with resolution, cracking problems in the heart of the cell also suffers from a lack of resolution in terms of how much we can actually see at the atomic level; even our best techniques like NMR spectroscopy and x-ray crystallography are acutely limited with limited to both resolution and dynamics. And just like a drone can be stolen or feared, the complex machinery inside a cell can interact very unpredictably with an intruder from outside, like a small molecule drug; it can chew up the drug or turn it into something toxic. Finally, the regulatory hurdles that drugs have to face are orders of magnitude bigger than those faced by drones.

There have certainly been honest attempts to tackle the complexity of biology recently, most significantly through machine learning and simulation approaches. But what the problems with drone delivery indicate is that some humility is in order here: one would have thought that Amazon Drone Delivery would have been right on the heels of Amazon Prime 1-Day Delivery. The basic issue is the distinction between code and the physical world. Code is clearly human created, cities are community created and bodies are crafted by four billion years of evolution. With code you know exactly where everywhere is, and there are clearly well known guidelines for debugging. If you don't like it you can redesign it from the ground up; try doing that with either cities or flesh and blood. In case of cities and even more so in case of biology, we don't even know where the bugs are, let alone how to debug them. It's a brave new world where we very much make the rules as we go along.

Clearly the drone delivery goal turned out to be deceptively simple, and the idea of applying software to drug discovery and biochemistry will be even more so. That does not mean progress won't be made (in both drone delivery and computational biology) and it certainly does not mean that software engineers should give up on trying to "solve" drug discovery, but it does mean that they need to be in for a long haul filled with blind alleys, sunk capital and plenty of heartache.

As the article says, coders who moved from the messy world of atoms into the clean world of code are now being confronted with addressing the messy, daunting world of atoms again. And there is no assembly of atoms conceivably more complicated that the one typing these words. From one flying jumble of atoms to another typing jumble, Palo Alto has a long way to go and I wish them luck.