This seems closely related to "the bikeshed effect" or "bikeshedding", in which a reviewer of a plan has more to say about the trivial (but easily understood) parts of the plan, than about the more essential (but difficult) ones. Everyone can have an opinion about what color the bikeshed at the nuclear power plant should be painted, but few spend the effort to understand the design of the reactor.
This is a linkpost for https://lydianottingham.substack.com/beware-the-delmore-effect
You should be aware of the Delmore Effect:
“The tendency to provide more articulate and explicit goals for lower priority areas of our lives.” —Cognitive Bias Foundation
It seemed responsible when I first encountered AI’s importance not to swallow it hook-line-and-sinker. So I did a shallow pass over neurotech, kept reading philosophy, and loaded up my math degree with physics-related (differential equations) courses as soon as I got the chance, with a view to understanding unrelated modelling, chaos theory, and so on.
I won’t argue these things don’t matter (in general and for AI), but in my case, they were trading off against higher-priority ML fundamentals: linear algebra, multivariable calculus, probability theory, CS, and programming.
I used to be confused about people who had what seemed to me like low-dimensional identities focused on popular, anti-esoteric topics. If you’re into AI, math, philosophy, and startups—who isn’t? It seemed somehow gauche or whatever. Some contrarian instinct in me wanted to escape it. Now I think it’s fantastic. By all means, go straight for what matters most.
It’s not just a matter of field selection. Nicholas D points out that a great paper has higher returns than a great blogpost, but writing a blogpost is easier, so he does it more often. Analogously, I’ve distracted myself with administrative homeostasis over work that might last for years.
Strive to swerve the Delmore Effect, and I think you’ll be better off for it.