Researchers at the University of Pennsylvania and Osmo have developed an AI system that outperforms humans in identifying odors. The system, trained on the molecular structure of 5,000 odorants, can describe what a molecule should smell like in human language. It also successfully mapped 500,000 odor molecules that have never been synthesized, a task that would take a human 70 years. This could advance understanding of how our brains and noses work together, and aid in the development of better mosquito repellents or deodorizing products.