Fermi estimates or problems are quick heuristic solutions to apprently insoluble quantitative problems rewarding clever use of real-world knowledge and critical thinking; bibliography of some examples.
2019-03-29–2021-06-11 in progress certainty: possible importance: 4
A short discussion of “Fermi calculations”: quick-and-dirty approximate answers to quantitative questions which prize cleverness in exploiting implications of common knowledge or basic principles in given reasonable answers to apparently unanswerable questions. Links to discussions of Fermi estimates, and a list of some Fermi estimates I’ve done.
Not only are they fun to think about, they can be amazingly accurate, and are extremely cheap to do—because they are so easy, you do them in all sorts of situations you wouldn’t do a ‘real’ estimate for, and are a fun part of a physics education. The common distaste for them baffles me; even if you never work through Hubbard’s How to Measure Anything (some strategies) or Street-Fighting Mathematics or read Douglas Hofstadter’s 1982 essay “On Number Numbness” (collected in Metamagical Themas), it’s something you can teach yourself by asking, what information is public available, what can I compare this too, how can I put various boundaries around the true answers2 You especially want to do Fermi calculations in areas where the data is unavailable; I wind up pondering such areas frequently:
- is a lip-reading website a good idea?
- how many women dye their hair blonde
- how many people does Folding@home kill
- what’s the entropy of natural language
- or how big a computer is needed to compute the universe
- do men have shorter real lives than women
- how much do Girl Scouts cookies cost and earn them
- how many people have used modafinil, and what is the cheapest we can expect to find modafinil for
- quickly assessing the rough probability of some event as I made one of my thousands of predictions
- how many sellers are there on Silk Road 1
- checking whether posthumous organ donation justifies the Chinese justice system (no)
- how many tourists to the Egyptian pyramids there have been compared to the workers who build them
- what we can infer from adultery & false paternity rates
- how many times have video gamers killed Mario
- is tiger skin more profitable than tiger bone wine?
- plating Versailles with gold
- the number of people killed by falling pianos
- the plausibility of _Oreimo’s Kirino Kosaka character
- how many integers one to a million do you see over a lifetime? (R simulation + Benford’s law)
- how many anecdotes of CrossFit causing rhabdomyolysis does it take to justify “CrossFit causes rhabdomyolysis”?
- how long would it take to read every book in existence?
- what would it cost to replace the US nuclear arsenal?
- What are the lifetime odds of being pooped on by a bird?
To look further afield, here’s a quick and nifty application by investor John Hempton to the Sino Forestry fraud: “Risk management and sounding crazy”. What I found most interesting about this post was not the overall theme that the whistleblowers were discounted before and after they were proven right (we see this in many bubbles, for example, the housing bubble), but how one could use a sort of Outside View/
Simple questions and reasoning can tell us a lot.
This is a little misleading; dimensional analysis is much more like type-checking a program in a language with a good type system like Haskell. Given certain data types as inputs and certain allowed transformations on those data types, what data types must be the resulting output? But the analogy is still useful.↩︎
eg. if someone asks you how many piano tuners there are in Chicago, don’t look blank, start thinking! ‘Well, there must be fewer than 7 billion, because the human race isn’t made of piano tuners, and likewise fewer than 300 million (the population of the United States), and heck, Wikipedia says Chicago has only 2.6 million people and piano tuners are rare, so there must be many fewer than that…’ You always know something, and have a universe of beliefs which imply constraints on everything and equilibria.↩︎