Researchers at UCLA have developed an optical generative AI model that creates images using minimal energy and computational steps.
Key Details
- 1UCLA Samueli School of Engineering introduces a photonics-based generative AI system.
- 2The optical approach generates images in a single step, compared to hundreds or thousands for digital models.
- 3Experimental and numerical tests showed high-quality image generation on datasets including human faces and Van Gogh-inspired artwork.
- 4The system incorporates a physical key-lock mechanism for enhanced image security and privacy.
- 5Potential applications identified: biomedical imaging, diagnostics, augmented/virtual reality, and portable/wearable devices.
- 6The approach significantly lowers energy and water consumption compared to conventional generative AI.
Why It Matters
This breakthrough may enable sustainable, fast, and secure generative AI with applications in medical imaging, diagnostics, and beyond. Its reduced resource requirements are particularly relevant as the environmental impact of AI computing grows.
Source
EurekAlert
Related News
•EurekAlert
AI-Powered OCT Enables Rapid 'Optical Biopsy' for Early Endometrial Cancer Detection
A team at Washington University has developed a catheter-based 3D OCT system with AI to quickly and noninvasively detect early endometrial cancers.
•EurekAlert
AI Clinical Reasoning in Diagnostics and Digital Fatigue in Healthcare
Recent JMIR features explore large language models in clinical diagnostics and digital fatigue among healthcare professionals.
•EurekAlert
KAIST, MIT, Microsoft Develop Efficient AI Image Upsampling for Robotics
KAIST, MIT, and Microsoft have created 'Upsample Anything,' a training-free AI method to restore high-resolution visual data from compressed images with up to 16x improved GPU memory efficiency.