- AI is evolving from a tool to a core part of scientific infrastructure, enabling smaller teams to handle tasks like literature review and experimental design.
- The focus has shifted from AI increasing science production to how it reshapes the questions scientists ask, with studies showing higher paper and citation rates but reduced topic diversity and collaboration.
- AI risks narrowing scientific inquiry and reasoning styles, exemplified in fields like depression research where it can bridge disciplinary gaps but may foster automation over deep investigation.
- Uncritical AI adoption in science is alarming, necessitating guardrails to prevent industrialization of research through automated pipelines that produce similar, publishable studies.