A California high school student, Matteo Paz, has stunned the scientific community after using artificial intelligence to identify around 1.5 million previously unknown objects hidden within old NASA space data. The breakthrough came from reanalyzing infrared observations collected by NASA’s NEOWISE mission, which originally focused on tracking asteroids and near-Earth objects. Scientists believed the massive archive had already been thoroughly studied, but Paz’s AI system uncovered patterns and variable objects that had gone unnoticed for years.
Paz developed the machine-learning system during a research program connected to the California Institute of Technology. Working with astronomers and mentors, he created an algorithm called VARnet that could rapidly analyze enormous datasets containing nearly 200 billion measurements gathered over more than a decade. The AI was trained to detect tiny fluctuations in infrared light, allowing it to identify variable cosmic objects such as quasars, exploding stars, eclipsing binaries, and possible black holes.
The discoveries were significant enough to earn Paz major recognition in the scientific world. His research was published in The Astronomical Journal, and he later won the prestigious Regeneron Science Talent Search along with a $250,000 prize. Researchers praised the project not only for the number of discoveries but also for demonstrating how artificial intelligence can unlock hidden insights from archival scientific data that humans alone would struggle to process efficiently.
Experts say the achievement highlights how AI is transforming modern astronomy and scientific research more broadly. Instead of relying solely on expensive new telescopes or missions, scientists can now revisit existing datasets using advanced machine-learning techniques to uncover discoveries that were previously impossible to detect. Paz himself noted that similar AI systems could eventually help analyze Earth-based problems as well, including environmental and atmospheric data patterns such as pollution cycles and climate trends.
