AI Chip Enhances X-ray Imaging Efficiency at Powerful Lasers

Scientists at SLAC National Accelerator Laboratory have developed a compact AI chip that significantly reduces the X-ray imaging bottleneck at high-power lasers. This innovation integrates intelligence directly onto a chip next to the detector, enabling inference in just 45 microseconds with minimal power consumption. The study, published in Frontiers in High Performance Computing, details how a specialized SpeckleNN neural network, optimized for a field-programmable gate array (FPGA), maintains 90% classification accuracy while achieving a 98.8% reduction in parameters, accelerating scientific research.