Anthropic publishes guest post on AI-assisted scientific modeling

AIAnthropic's Science Blog publishes a guest post by Jakob Macke, Daniel Gedon, and Stefan Wahl on how AI is starting to help with writing, fitting, and revising scientific models. The authors say that by treating models as code, AI can explore many possible model structures and show which explanations the data support.

AIA guest post by ELLIS Institute Tübingen researchers Jakob Macke, Daniel Gedon, and Stefan Wahl says AI is starting to change scientific modeling. They say AI systems can now help build models and constrain parameters, using simulation-based inference and LLM agents that propose candidate models. The authors argue that interpretation remains hardest to hand to AI, and that returning a range of plausible models gives more insight than a single fit.