Scientists at Oak Ridge National Laboratory are developing AI-enabled pixel detectors that can analyze particle-collision data directly at the source. The approach could help particle-physics experiments identify and capture the most important signals from the enormous amounts of data modern accelerators produce, helping scientists make faster, more informed discoveries from some of the world's most complex experiments.
The project, called NEUROPix - short for neuromorphic computing for pixel detectors - recently received a three-year award through the Department of Energy's High Energy Physics program. The funding supports efforts to use artificial intelligence directly within scientific instruments to process data in real time.
The ORNL team will use spiking neural networks, a form of neuromorphic computing inspired by the human brain, to identify patterns and extract valuable signatures from particle interactions in real time - an approach that could benefit many other data-intensive scientific instruments.
"Our particle accelerators can now generate much more data than we're able to record to disk," said ORNL physicist Mathieu Benoit. "The idea is to deploy intelligence close to the detector so we can sort or compress the data very quickly while keeping the information that matters most."
- Galen Fader
