We hope your summer is off to a great start. For our new readers, this might be the first edition of Hold the Code youâre readingâwelcome! During the summer, HTC will release on a less frequent schedule and experiment with different formats and content. Stay tuned for whatâs coming and tell a few friends to subscribe!
NASA released the first full-color images and data snapped by the James Webb Space Telescopeâs (JWST), which includes the deepest and sharpest infrared image of the distant universe to date. The largest and most powerful space observatory ever built captured a photo of galaxy cluster SMAC S0730, dubbed as Webbâs First Deep Field. AI already has a role in analyzing data from the telescope.
Whatâs the big deal about the photos?
The image of Webbâs First Deep Field reveals thousands of galaxies in a miniscule sliver of the universe. It also offers the most detailed view of the early universe so far. NASA Administrator Bill Nelson said the telescope can see back almost to the beginning of the universe and aid in determining if planets outside our solar system are habitable
âWe are going to be able to answer questions that we donât even know what the questions are yet,â said Nelson.
Created at UC Santa Cruz (UCSC), a new generation of GPU-accelerated AI will help scientists ask and answer these questions.
Where does AI come in?
Brant Robertson, an astrophysics professor at UCSC helped develop Morpheus, a machine learning model that uses a deep learning framework to classify galaxies and other astronomical objects by poring over pixels and analyzing raw data coming from the telescope.
Robertson said he believes Morpheus will lead to breakthroughs that better inform us of how the universe formed almost 14 billion years ago.
Morpheus previously helped scientists understand images taken from NASAâs Hubble Space Telescope. Robertson and his team have updated the algorithm, which can now:
classify larger images roughly 100 times faster
use advanced image processing capabilities to separate objects that appear to overlap in the sky
What's Next?
Initially trained on 7,600 images taken from the Hubble telescope, Morpheus still requires more training. The JWST will collect data from light galaxies much further away than Hubble could see (see the differences here). Robertson said what a star looks like through the JWST drastically differs.
Morpheus runs on UCSCâs Lux supercomputer that includes 28 GPU nodes with two NVIDIA V100 Tensor Core GPUs each.
"Once the data is in hand, running Morpheus on all the JWST images will only take a few days at most on lux," Robertson said.
Take a closer look
Curious about Webbâs First Deep Field? UCSC made the image in a zoomable format so you can take a closer look at the galaxies.
Love HTC? â€ïž
Follow RAISO (our parent org) on social media for more updates, discussions, and events!
Instagram: @Raisogram
Twitter: @raisotweets
RAISO Website: https://www.raiso.org