NASA has announced the anticipated launch of the Nancy Grace Roman Space Telescope, set to take place in September 2026—eight months earlier than previously scheduled. This new telescope is expected to yield an impressive 20,000 terabytes of data throughout its operational lifespan.
In comparison, the James Webb Space Telescope has been providing around 57 gigabytes of striking imagery daily since its inception in 2021. Additionally, the Vera C. Rubin Observatory in Chile will commence its survey later this year, aiming to gather 20 terabytes of data every night.
For context, the once-revered Hubble Space Telescope delivers only 1 to 2 gigabytes of data on a daily basis. Typically, astronomers have not been able to manually review all this data and are now increasingly employing GPU technology to assist in their analyses.
Brant Robertson, an astrophysicist from UC Santa Cruz, has been instrumental in leveraging GPU technology to tackle challenges in space science. Over the past 15 years, in collaboration with Nvidia, he has focused on advanced simulations to assess theories about supernova explosions and created analytical tools for processing substantial datasets from contemporary observatories.
Robertson explained the evolution in methodologies: “There’s been this evolution [from] looking at a few objects to CPU-based analyses on large datasets, progressing to GPU-accelerated analyses.”
Alongside his then-graduate student Ryan Hausen, he developed Morpheus, a deep learning model that efficiently scans extensive datasets for galaxy identification. Their initial analyses using Webb data revealed an unexpected prevalence of certain disc galaxies, potentially reshaping existing theories on universal development.
To remain at the forefront of advancements, Robertson is transitioning Morpheus from convolutional neural networks to transformer models, akin to those used in recent large language models. This shift is expected to significantly enhance the model’s analytical capacity and speed.
Moreover, Robertson is innovating generative AI models using telescope data to enhance the quality of observations gathered by ground-based telescopes, which are affected by atmospheric distortion. As deploying large mirrors into orbit remains challenging, this software-based approach provides a valuable alternative.
Despite progress in technology, there remains a pressing demand for GPU access among researchers. Robertson has established a GPU cluster at UC Santa Cruz with support from the National Science Foundation. However, with increasing demand for compute-intensive techniques amidst ominous budget cuts proposed by the Trump administration, there are concerns about resource availability.
Robertson remarked, “People want to do these AI, ML analyses, and GPUs are really the way to do that. You have to be entrepreneurial…especially at the edge of where the technology is.” His experience underscores the need for academic institutions to embrace innovation while navigating constrained resources.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
