Artificial intelligence is helping scientists find inaccuracies in old scientific works. For instance, a model discovered incorrect data in a chemistry reference book that had persisted for 75 years, reports infohub.kz.
Researchers have begun using AI agents to analyze scientific literature and databases. The models successfully identify errors and typos that have lingered for decades in authoritative sources.
Theoretical chemist Sebastian Pios from Zhejiang Lab in Hangzhou used AI to predict boiling points of molecules. When the model showed a result differing from the known reference, the scientist suspected an AI error. However, checking primary sources proved that the incorrect data had been in the reference book for 75 years. AI also helped find typos and inaccuracies in century-old measurements.
Reference constants form the basis for educational problems, industrial calculations, and training new algorithms. Therefore, automated literature checking becomes an important audit tool. Researchers emphasize that AI itself can make mistakes, so all found discrepancies require manual verification by experts.