(Audio) Loonshots, by Bahcall

Monday July 22, 2024

It's a fun book, especially at the beginning, where Bahcall has a lot of underdog success stories. I think some of the ideas are decent, but it's also more like an unfinished brainstorming session than a settled science, and it's hard to distinguish useful theory from just-so history. I'm fundamentally on board for more discoveries though!

A major metaphor in the book is "phase transitions," like solid to liquid. Bahcall offers "The Innovation Equation" (details in HBR article) for \( M \), "a certain size at which human groups shift from embracing radical ideas to quashing them."

\[ M = \frac{E S^{2} F}{G} \]

The only right hand term that isn't unitless is \( S \), "management span," which is the average number of direct reports that managers have. Since \( S \) is squared, the result is in units of people squared, which I don't think makes sense. With his training as a physicist, it's a little surprising that Bahcall would do this. (Am I missing something?)

Also surprising is that there's no constant term. (For example, gravity isn't mass times mass divided by distance squared; there's a constant that we have to get from data.) Bahcall plugs in \( E = 50\% \), meaning compensation is 50% equity, which seems rare, then management span of 6, and \( F = 1 \) meaning equal returns to project work and politics, finishing with \( G = 12\% \) meaning 12% compensation increase at promotion. Ignoring the units, this gives 150, and he claims a connection with Dunbar's number, saying "That's interesting."

I have to think Bahcall doesn't really take this equation seriously; he must intend it as an illustration for his narrative. Still, it offends my sensibilities because it seems like math-washing: presenting something that looks "fancy" to make it seem more legitimate. In this case, the evidence isn't really anything more than suggestive.

One issue is boundaries: Where does an organization begin and end? DARPA is considered an example of an innovative organization. Do we consider just the DARPA PMs? Also their contractors? The whole Department of Defense?

I think a much more promising attempt might be made by switching from the organization to the individual view: At what point does an individual change their behavior, based on the local structure of their organization and incentives? This would also be more coherent with Bahcall's ideas about having different parts of an organization working on different goals (loonshots vs. "franchise" work). How do you get loonshots? Support people who work on them.

A deeper critique is that Bahcall's "loonshot" forces here seem more about doing your job than about working on crazy innovative ideas specifically. "Project work" is not necessarily loonshot work.

This "Goals Gone Wild" paper is pretty fun, and reminds me of The Tyranny of Metrics. The paper includes this warning label:

Also referenced (a lot) is Vannevar Bush's 1945 Science: The Endless Frontier, which reminds me of The Beginning of Infinity.

“The graveyard of unexplained experiments, as Land would soon show, is a great place to find a False Fail.” (page 76)

“Some companies are the equivalent of an innovation landfill,” wrote one senior Apple executive, who helped lure some of PARC’s best engineers to Apple. “They are garbage dumps where great ideas go to die. At PARC, the key development people kept leaving because they never saw their products get into the market.” (page 115)

This is an interesting idea: What's out there that's known but ignored, currently, when it shouldn't be ignored? Bahcall uses the example of radar's foundations being known but ignored by the Navy for nearly 20 years. What fun things are out there, already discovered but not yet exposed to sunlight?

“While the individual man is an insoluble puzzle, in the aggregate he becomes a mathematical certainty.” (Sherlock Holmes, in The Sign of the Four, quoted page 117)

Seems like everybody wants to do psychohistory!