The Spectrum Dispatch News

movies

Examining Motivation Limits in Scientific Discovery: Insights from Gwern’s 'On Really try'

The essay argues that knowing a theory is wrong does not guarantee the drive to replace it, citing historical examples and anecdotes from cryptography and puzzles.

Examining Motivation Limits in Scientific Discovery: Insights from Gwern’s 'On Really try'

The piece opens with a quote from Zen teacher Yasutani: conviction creates determination, which in turn creates zeal, and without conviction one cannot awaken no matter how much practice is undertaken. The author connects this to Eliezer Yudkowsky’s ‘beisutsukai’ post on the failures of ‘Eld Science’, noting that even when scientists intellectually know a theory is wrong, they may lack the psychological energy to pursue a replacement unless they feel an immediate need for an answer. Jeffreyssai, a fictional instructor, argues that Eld scientists treated a thirty‑year timeline as acceptable because their social process assumed truth would eventually emerge; work expands to fill the time allotted. He contrasts this with the urgency of dodging a thrown knife, suggesting that expectations of speed can dramatically shorten problem‑solving time. The text highlights a psychological advantage for learners who start from a known false theory: they are more willing to discard painful insights and avoid spending effort patching the old model. The author adds a personal observation that motivation rises when one is assured a correct answer exists, compared to uncertainty about whether a question is answerable. The argument is that breakthroughs depend less on innate competency at paradigm shifts and more on accepting that a shift must happen now and that one is responsible for making it. A classroom scenario is described in which students are tasked with producing a correct theory of quantum gravity within one month; the instructor says that without knowing the answer they are not expected to solve it unless they retain a partial hint. An anecdote about Claude Shannon illustrates how a minimal hint—his brother saying “I could tell you something”—was enough to solve a jigsaw puzzle, and that such hints can be self‑generated. The essay also references Manuel Blum’s advice to beginning graduate students, though the exact wording is not reproduced. A Kaggle observation is noted: after a long leaderboard stagnation, a single team’s breakthrough often triggers multiple independent teams to reproduce the result quickly, even without knowing the first team’s method. Finally, the narrative recounts Ralph Merkle’s work on secure communication over a channel with an unknown partner. Merkle submitted a term‑paper proposal that his professor Hoffman did not understand, leading to a rewrite request; Merkle left the class and later published the paper “Secure communication over insecure channels” about five years later. The delay in recognition is attributed, in part, to Merkle’s poor writing style, with a draft described as a hundred‑page manuscript that was hard to read.

Examining Motivation Limits in Scientific Discovery: Insights from Gwern’s ‘On Really try’

Key facts

  • Yasutani: conviction → determination → zeal; without conviction no awakening despite practice.
  • Jeffreyssai: Eld scientists accepted 30‑year timelines because they expected truth eventually; work expands to fill allotted time.
  • Known false theory gives willingness to discard old insights and avoid patching.
  • Personal note: more motivated when assured a correct answer exists than when unsure question is answerable.
  • Class Project: one‑month limit to produce quantum gravity theory; without answer not expected to solve unless retaining a hint.
  • Claude Shannon anecdote: brother’s hint “I could tell you something” enabled solving a jigsaw puzzle; hint can be self‑given.
  • Kaggle fact: after leaderboard stagnation, a breakthrough prompts multiple teams to reproduce it quickly without knowing the first team’s method.
  • Ralph Merkle: submitted term‑paper proposal on key exchange; professor confused; Merkle left class; later published ‘Secure communication over insecure channels’ ~5 years later; delay partly due to poor writing (hundred‑page hard‑to‑read draft).

Sources

← All posts