Ethan Mollick argues AI could change how science picks research questions

Ethan Mollick says on X that academics invest heavily in each paper, so they tend to pursue predictable work likely to yield significant results.

He argues AI faces no such constraint, which could force a rethink of how science is done. Mollick's post cites a working paper by Eva Vivalt on whether AI can forecast economic treatment effects from pre-registration documents, and reports that the answer is yes.