Exploring AI’s Search for Extraterrestrial Life

Andrew Siemion, the Bernard M. Oliver Chair for SETI at the SETI Institute, expressed excitement about the future of analyzing streaming astronomical data. Siemion highlighted the potential for groundbreaking discoveries with the evolving technology. The SETI Institute, established in 1984 and now comprising over 120 scientists, operates the cutting-edge Allen Telescope Array in Northern California. This telescope is instrumental in the search for extraterrestrial intelligence (SETI) and the study of transient astronomical events like fast radio bursts.

The inception of the latest project dates back over a decade when Siemion attended a talk at the University of California, Berkeley, on early machine learning classifiers. This event sparked a realization that the methods used by SETI researchers to detect signals were relatively basic. Siemion, who holds a Ph.D. in astrophysics from Berkeley, began collaborating with radio experts outside the astronomy field to explore new approaches.

One key connection was made with Adam Thompson, who leads a group of developers at NVIDIA. The SETI Institute researchers had been utilizing NVIDIA GPUs to enhance signal separation algorithms for years. The potential for further innovation became apparent, leading to discussions on leveraging NVIDIA technology to address the challenges of processing wide bandwidth signals from space at high data rates.

Despite delays caused by the pandemic, Thompson eventually introduced Siemion's team to NVIDIA Holoscan earlier this year. Holoscan, a sensor processing platform for real-time data from scientific instruments, caught the attention of the researchers. They decided to develop a trial application using Holoscan on the NVIDIA IGX edge computing platform, aiming to revolutionize the SETI Institute's operations.

The collaboration extended to Breakthrough Listen, a SETI research program based at the University of Oxford. This program, utilizing numerous radio telescopes to collect vast amounts of data, collaborates with the SETI Institute. The joint effort aimed to create a streaming solution that could process real-time telescope data directly on GPUs for AI inference, enhancing the capabilities of both organizations.

Luigi Cruz, a staff engineer at the SETI Institute, spearheaded the development of the real-time data reception and inference pipeline using the Holoscan SDK. Meanwhile, Peter Ma from Breakthrough Listen focused on building and training an AI model to detect fast radio bursts. Wael Farah, the Allen Telescope Array project scientist, provided crucial scientific insights for the study.