Rocks over 3 billion years old have undergone geological changes like burial, compression, and high temperatures, losing all intact fossil traces. How to find life evidence in them? A team from the Carnegie Institution for Science made a breakthrough with AI technology—by decoding faint "chemical echoes" in rocks, they not only discovered life traces in 3.3-billion-year-old rocks but also rewrote the evolution timeline of photosynthesis, even offering a new path for finding life on Mars.

Key Findings

1. Old methods fail, AI solves "chemical noise" problem
   When studying rocks older than 1.7 billion years, traditional chemical analysis could only deal with chaotic molecular data (mostly "chemical noise" from geological activities) and couldn’t distinguish life traces. The team first crushed rock samples into molecular fragments with advanced technology, then used supervised AI to filter signals—feeding the AI over 400 samples (including modern organisms, young fossils, meteorites, and other non-biological substances) to teach it to distinguish "biological chemical fingerprints" from "non-biological chemical signals," finally achieving an accuracy rate of over 90%.
2. Life evidence from 3.3 billion years ago, doubling the research time window
   AI identified subtle chemical structures formed only by life in 3.3-billion-year-old rocks. This discovery pushes the detectable life timeline from 1.7 billion years ago to 3.3 billion years ago, opening a brand-new window for studying early Earth life.
3. Photosynthesis evolution advanced: Oxygen-producing mechanism existed 2.5 billion years ago
   AI also found key clues: chemical traces of oxygen-producing photosynthesis in 2.5-billion-year-old rocks. This subverts traditional cognition—previously, the "Great Oxidation Event" (when atmospheric oxygen surged) was thought to occur 2.4 billion years ago, but the evolution of oxygen-producing photosynthesis may be hundreds of millions of years earlier. At that time, microorganisms may have already formed localized oxygen-rich environments.
4. Seaweed fossil verification: AI resists "time damage"
   To ensure reliability, the team validated the method with 1-billion-year-old seaweed fossils. Despite billions of years of geological activity breaking their chemical structures, AI could still accurately identify their biological properties, proving the technology can withstand "time damage" to ancient rocks and the results are credible.
5. New possibility for finding life on Mars: Reread existing data
   Mars has no plate tectonics, so ancient life traces in its rocks may be better preserved than on Earth. More importantly, NASA’s Mars rovers have collected a lot of chemical data, which could not be interpreted by old methods. Reanalyzing this data with this AI technology may allow us to find evidence of ancient Martian life without launching new rovers.