Across the pipeline, the work splits along a clear line. AI owns the fast, repetitive, well-specified inner loop: predicting properties, generating candidates, ranking which experiment to run next, and running the cycle again. Humans own the framing and the judgment points: choosing the problem, deciding what an ambiguous result means, and answering for safety and patient outcomes. This is not a temporary division that better models will erase; it follows from what each side is actually good at and what each can be held responsible for.
Can AI Run a Drug Discovery Lab on Its Own?
What Still Needs a Human
Who Owns Which Step
3 / 4
So the loop pauses for a person, then resumes. Zoom out across the whole pipeline, and a pattern shows up: the work splits along a fairly clean line.
0:00 / 0:00