Concept · part of Thinking inside frameworks

Write 1,968 as MCMLXVIII and try to multiply it by 47 on paper. Now write it as 1968. The second version lines the digits up by place, and that is what makes written multiplication a routine. Roman numerals were fine for recording a result and poor for reaching one.

Kenneth Iverson opened his 1979 Turing Award lecture with the point that notation is a tool of thought. A good notation, he argued, must make it convenient to express the ideas that come up later, when you analyze and generalize, as well as the first ones. Leibniz’s way of writing calculus does this and Newton’s does less of it, and British mathematics, which kept Newton’s, lagged behind the Continent (calculus-notation, with the caveat that notation was one cause among several). Programming languages work the same way: each makes certain solutions feel natural.

A notation is a small, clean framework, and cost asymmetry with little else mixed in. It is also the hopeful case. A good notation lets a finite mind hold more, which is the job the book hopes a faithful AI tutor could do. The catch is lock-in: people keep the notation they were trained in.

Read the full argument

How it connects


← The Columbia slide · Thinking inside frameworks (6 of 21) · Leibniz’s dy/dx and the British lag →

Explore all the ideas · Read the book in order · Open the interactive map