GoKawiilUCLA builds optical AI system that screens multiple deepfake videos at once
UCLA researchers led by Professor Aydogan Ozcan built a hybrid digital-optical neural processor that detects deepfake videos by using light to analyze 15 or more video streams simultaneously in a single optical pass. The system, described in a study published in eLight, pairs a lightweight digital encoder that extracts spatial, spectral and temporal features with optical processing to flag manipulated footage.
GoKawiil's interpretation of the reporting above, not reported fact.
Conventional deepfake detectors typically process videos one at a time and can require hundreds of billions of computations per analysis, making large-scale screening slow and energy-intensive. By performing part of the detection optically and in parallel, this approach could reduce processing time and energy use while serving as an early screening layer, though its real-world resilience against adversarial manipulation still needs broader testing beyond the lab study.
- UCLA's optical-neural processor can analyze 15+ video streams simultaneously using light-based computation.
- The hybrid system combines a digital encoder for feature extraction with optical processing for speed and efficiency.
- Researchers position it as a first-layer, high-throughput defense against large volumes of AI-generated video.
Source: sciencedaily.com, 2026-10-01
Published there as: “This light-powered AI can spot deepfakes with nearly 98% accuracy”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.