Hypotheses

Scientists are notorious for their disdain toward new proposals (ideas, explanations, connections), especially when they come from outside science. This deprecation has nothing to do with scientific method, which makes no demands on hypothesizers. Experimentation requires rigor, and conclusions require training, but hypothesizing actually benefits from a dilatory big-picture—even poetical—approach.

Hypothesizing can't be trained. It's a creative flexion for which some people are more suited than others. And the precise, linear style of thought of those who choose careers in science is not known for fostering creativity and insight. Scientists can be outstanding hypothesizers, but it's despite their training and milieu, not because of it. Many are conservative to the point of hidebound.

Just as it's risible that politicians are expected to not just garner votes but also run things, it's odd that we expect scientists to dream up hypotheses. A poet—anyone versed in disciplined dreamy speculation—might be better suited.

This exclusion has been willful but made necessary by limited bandwidth and poor signal-to-noise. It would be impossible to triage (much less test and prove) every daft notion streaming in from outsiders. But a poor signal-to-noise ratio does not augur a low ceiling. The lost gems might have been immensely useful. Some people are immensely creative and insightful, and most of them don't go into science, so their contribution is lost.

LLMs could perform this triage at scale, uncomplainingly, with deep knowledge and institutional skepticism approximating a trained scientist. Such hypothesis mining could make a profound impact.