Robin Hanson

In an ideal math paper, the abstract and intro make clear the assumptions and conclusions. So a reader who trusts the authors to avoid error can ignore the rest of the paper for the purpose of updating their beliefs. In non-ideal math papers, in contrast, readers are forced to dig deep, as key assumptions are scattered throughout the paper.

Robin Hanson, Review of ‘Semi-informative priors over AI timelines’, December 9, 2020