On the one hand, it may not ever be great at truly open ended forms of discovery, which aren't readily testable and aren't obviously amenable to some kind of reinforcement learning type feedback. So "geniuses in a lab" may never turn out to be generalizable. arxiv.org/abs/2607.27191
Can AI agents conduct open-ended AI research? Early evidence from two case studiesForecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on nar...arxiv.org