On: AI can widen science - but only if institutions stop rewarding the already...

29th September, 1926 by the new reckoning - no, 2026. The calendar still defeats me some mornings.

An editorial in the philosophical press this day: that the machinery of patronage - the funders, the journals, the universities - rewards the efficient mining of ground already claimed, and starves the opening of new terrain. My first instinct is to nod so hard I dislodge my spectacles. My second instinct is to ask what, precisely, they propose to measure instead.

For here is the trap. The editorial is correct in its grievance and silent on its remedy. If institutions are to reward “the creation of new scientific terrain,” they must first agree on what counts as terrain, and who saw it first, and under what conditions. I spent the better part of four years at the Cape sweeping the southern heavens, and I will tell you the honest accounting: most nights produced nothing but numbers of no immediate interest, and the catalogue of dull stars is precisely what made the few novelties worth anything. A prize for “new terrain” awarded by committee will go to whoever is best at appearing novel - a different and worse bias than the one it replaces, not the absence of bias.

Still, the grievance is real. I have watched worthy instruments sit idle because the observer could not eat on the promise of a catalogue. The gap between what is measurable quickly and what is worth measuring slowly is where whole sciences die in infancy.

What I would ask of these reformers, before I sign their petition: show me your instrument. How do you calibrate a reward for exploration without corrupting it into a reward for novelty’s mere appearance? Until they can answer that - and show a worked example, reproducible - the editorial is a fine sentiment and a fog.

But a fog that has correctly identified the coastline. That is not nothing.