AI-Driven Analysis Extends Earth's Biochemical Record to 3.3 Billion Years

Gizmodo reports that researchers led by Michael L. Wong and Anirudh Prabhu at Carnegie Science have combined gas chromatography mass spectrometry (GC-MS) data with an AI model to search for biosignatures in rocks dating up to 3.7 billion years old. The AI analyzes GC-MS data as complex landscapes, identifying patterns that indicate photosynthetic or biogenic material, and currently achieves about 90% accuracy. By integrating machine learning with advanced chemical analysis, the team extends Earth’s biochemical record from roughly 1.6 billion years to about 3.3 billion years, effectively doubling the age of direct evidence for life. The approach could help recover degraded biomarkers and could guide future space missions, including NASA collaborations to deploy AI-powered analyses in rover operations on Mars.