Daniel Lemire's blog

, 2 min read

Rigor or relevance: choose one

2 thoughts on “Rigor or relevance: choose one”

  1. Kevembuangga says:

    In AI both the “scruffy” and the “neats” failed.
    May be there is something else to care for, that would explain the poor progress in the field.
    It may also mean that neither the likes of Marcus Hutter nor the robotics tinkerers will “crack AI”.
    What could be missing?

  2. Peter Turney says:

    A scientist or mathematician may achieve relevance as a side-effect of aiming for rigour. It was once pointed out to me that many of the most cited results in math are lemmas, rather than theorems. (For non-mathematicians in the audience, a lemma is a kind of mini-theorem that is proven as a step along the road to the main theorem.) Some of the things that engineers do are “superstitions”, which are later shaved off by rigorous science. For example, the first refrigerators encased the cooling coils in a saltwater solution, in superstitious imitation of ice boxes, the preceding technology for cooling food. My point is that we need to alternate back and forth between rigour and relevance. First make it work (relevance), then clean it up (rigour).