AI-powered tools have proven to be valuable assets in research, and mathematics is no exception. Our PI Peter Pfaffelhuber explores an innovative workflow assisted by Anthropic’s Claude: much of the paper’s mathematical content emerged through a dialogue between him and AI. Claude also drafted large parts of the manuscript and contributed to the underlying simulation code, which played a key role in identifying the right conjecture.
An inherent risk to such an approach is the possibility of hallucinations, which has been addressed by systematically formalizing results in the interactive theorem prover Lean. Interested parties can therefore independently verify their work by accessing the resources in their public repository.
We believe this work offers a glimpse into future new developments for mathematical research.
Read the preprint here.
Institute of Medical Biometry and Statistics,
Faculty of Medicine and Medical Center –
University of Freiburg
