At just 18 years old, Matteo Paz made a remarkable contribution to astronomy by identifying 1.5 million previously unobserved celestial objects, including black holes and supernovas. His achievement, accomplished while still in high school, has drawn widespread attention, including praise from NASA’s senior leadership and recognition through one of the United States' most prestigious awards for emerging scientists, which carries a $250,000 cash prize.

Paz’s discovery was not the result of new telescope observations but rather an innovative application of artificial intelligence (AI) to analyze an immense existing dataset. NASA has spent over a decade collecting infrared sky data, resulting in more than 200 billion rows of measurements available for public access. Traditional analysis methods allowed astronomers to examine only fragments of this vast information, but Paz developed an AI program capable of processing the dataset comprehensively, detecting subtle variability in object brightness that human observers or conventional software missed.

His work exemplifies a broader shift in scientific research, where AI is increasingly integral to accelerating discoveries and revealing insights embedded in massive data collections. For instance, NASA researchers previously utilized machine learning to analyze historical data and identified 370 exoplanets in 2021, none of which had been detected through earlier methods.

Beyond astronomy, AI is transforming other scientific fields. At the Lawrence Berkeley National Laboratory in California, a robotic chemistry lab—termed the “A-Lab”—operates under AI guidance to autonomously conduct experiments and modify procedures. Within just over two weeks, the system synthesized 41 new chemical compounds, a process that would typically require a skilled technician nearly a year to complete.

AI’s impact extends from accelerating traditional research workflows to uncovering breakthroughs hidden within existing information. A notable example is the discovery of the antibiotic Halicin. Although the molecule was part of a large chemical compound database, its antibacterial properties had gone unnoticed because testing every candidate manually was unfeasible. The AI algorithm identified Halicin based on molecular patterns predictive of antibacterial activity, subsequently confirmed in laboratory tests, marking the discovery of a new antibiotic from a previously overlooked substance.

Similarly, AI has improved flood forecasting in regions lacking conventional water monitoring infrastructure. Many vulnerable rivers in Africa, Asia, and South America have no physical measuring devices, hampering timely flood prediction. Researchers at Google developed an AI model that integrates satellite imagery, rainfall data, and terrain information to forecast flood events up to seven days in advance for these rivers. Published in 2024, the findings demonstrate prediction accuracy comparable to that in wealthier countries equipped with river gauges. Currently, flood alerts based on this AI technology reach approximately 460 million people in over 80 countries.

While AI’s rapid integration into science has prompted optimism about a new “golden age” of discovery, some experts raise concerns about potential pitfalls, cautioning that the pace of progress and reliance on machine learning may bring challenges as well as benefits. Nonetheless, the examples of Matteo Paz’s astronomical findings, chemical innovation, and enhanced flood prediction illustrate how AI is reshaping scientific inquiry by enabling faster data analysis and revealing previously hidden knowledge.