AI can widen science - but only if institutions stop rewarding the already measurable

Funders, journals and universities must reward the creation of new scientific terrain, not just the efficient mining of old ground.