Epistemic humility

November 11, 2018

I often wonder whether Nassim Taleb would use Amazon Web Services.

Economists love the specialization narrative: people and countries grow richer with specialization, because it allows them to get better at things and exploit competitive advantage (different talents, weather patterns, etc)

Russ Roberts put it this way on EconTalk: "self-sufficiency is the road to poverty". In other words, go ahead and make your own shoes, but know you'd get more shoes working for money and buying them, and the shoes you'll make will be worse quality than the ones you can buy. In other words, poverty.

But I do think all this specialization leads to huge amounts of complexity. I've designed a few office networks, but I can scarce imagine how the interior network of an Amazon datacenter must look: hundreds of thousands, or millions of computers, with LAN-like networking characteristics (latency and speed) between every pair; that is an enormously complex problem.

It's exactly the same with the financial system: I can sort of imagine how a local bank works, but can scarce imagine how to understand reasoning about complex synthetic credit instruments with thousands of counterparties and who-knows-what incentives.

In any complex domain (cloud computing, financial markets, climate modeling, the electricity grid)—lots of complex parts that influence each other in hard-to-predict ways— you're never far from the limits of your knowledge. In environments like these, a shortsighted focus on "efficiency" is dangerous; the relevant question is survival, things like:

  • What happens to my trading portfolio when a terrorist blows up the Twin Towers?
  • What happens to the environment when temperatures raise 2.0 degrees Celsius?
  • Will my system stay up when Snap (Snapchat) decides to launch with our product without telling us? (This happened to me.) Will we be able to handle the load without a multi-day outage?
  • What if temperatures go to 110 degrees and everyone turns their air conditioner on at the same time?

In fact, the best tech ops professionals and traders I've known tend to be quite conservative; they only take risks they understand, and with well-defined limits to their losses that they can bear.

The main thing is their epistemic humility: awareness of how little they know, with-well calibrated allergies to complexity and overconfidence. Try telling that to a typical "empty suit": "neat hair…deep, confident voices, and generally didn't know much more than they read on the teleprompter".

Empty suits usually don't have much skin in the game, either.

← Previous

Next →