Discovering the Cosmos: How an American Student’s AI Unveiled 1.5 Million Hidden Space Objects

A high school student in California has made waves in the field of astronomy by creating an impressive artificial intelligence model that discovered 1.5 million space objects previously unknown to us. During his summer research at Caltech, Matteo Paz not only pushed scientific boundaries but also authored a peer-reviewed paper detailing his findings, published in The Astronomical Journal.

The journey began in the summer of 2022 when Paz joined Caltech’s Planet Finder Academy. This program, led by Professor Andrew Howard, aims to expose students to advanced astronomical research. Paz, under the guidance of senior scientist Davy Kirkpatrick at IPAC (Caltech’s Infrared Processing and Analysis Center), took on the immense task of analyzing data from the NASA telescope, NEOWISE. Initially designed to track near-Earth asteroids, NEOWISE collected a wealth of information about more distant cosmic bodies over a decade.

These distant objects, such as quasars and supernovae, showed varying brightness levels, presenting a significant opportunity for exploration.

With nearly 200 billion data points to sift through, the team initially considered a manual approach. However, Paz drew on his strong background in computer science and theoretical math. He realized that this was a perfect fit for an AI solution. Within six weeks, he devised a Fourier and wavelet-based machine learning model to scan the massive dataset and identify variable brightness in objects.

His hard work paid off quickly. The model began detecting subtle shifts in infrared light, hinting at new celestial activities. According to Phys.org, the model showed promise almost immediately. As Paz iterated on the design, its accuracy and efficiency improved, allowing for the identification of objects that traditional methods might have missed.

Paz’s collaborative environment at Caltech played a crucial role. He worked alongside researchers like Shoubaneh Hemmati and Daniel Masters, who provided insights on machine learning and astronomical detection. As he delved deeper, Paz identified limitations in NEOWISE’s data collection methods. Some phenomena changed too slowly or flashed too briefly for conventional analysis to capture. His AI model empowered researchers to flag these elusive variable stars, revealing new insights into their behavior.

Paz’s findings culminated in a detailed paper that laid groundwork for a complete catalog of detected variable objects, set for release in 2025. This catalog opens new doors for astronomers interested in studying the long-term evolution of stars and galaxies.

His work isn’t just confined to astronomy. Paz sees applications for this model beyond space research. “The model can be adapted for any time-based studies,” he noted. This adaptability could extend to analyzing stock market trends or tracking environmental changes, showcasing the versatility of machine learning.

Now, as a paid employee at Caltech, Paz continues to refine his AI model and mentor fellow students in the Planet Finder Academy. His journey highlights the impact of mentorship and innovative research methods. Kirkpatrick, who has nurtured Paz’s potential, reflects on the importance of fostering young talent: “If I see their potential, I want to make sure they are reaching it,” he says.

Matteo Paz’s story is a promising example of how young minds are pushing the frontiers of knowledge, blending artificial intelligence with astronomical research, and inspiring the next generation of scientists.

Source link