University of Bielefeld researchers use AI to decode alchemy

A research team led by Dr. Guillermo Restrepo at the University of Bielefeld is using artificial intelligence to decode centuries-old alchemical recipes. The project, titled “(Al-)Chemical Space In Silico” (AXIS), aims to identify forgotten chemical knowledge hidden within historical texts that are often written in allegorical or esoteric language.

Supported by a three-year, 400,000 euro grant from the US-based non-profit Schmidt Sciences, the team is working alongside Dr. Sarah Lang from the Max Planck Institute for the History of Science. The methodology involves using Large Language Models (LLMs) to distinguish actual chemical content from symbolic descriptions.

Once potential recipes are identified, researchers will use quantum chemistry and computational methods to evaluate their scientific viability before attempting to verify the findings in a laboratory setting.