In this invited talk, I will review our recent theoretical, and experimental results in the field of artificial intelligence (AI) assisted inverse design of a new class of flexible ultra-flat optics (thickness < 100nm) for high efficiency (close to unitary) vectorial control of light. These devices are supported by a layer of “physical” neural network units in suitably engineered optical nanoresonators. These systems act as universal approximators of arbitrary defined input-output responses, processing information at the speed of light.
I will discuss basic optical components and new integrated systems for processing high-dimensional visual information in real-time and a new concept of universal integrated, ultra-flat equipment for metrology applications based on data-driven science. Finally, I will also discuss some recent challenges in manufacturing high-resolution, large-scale samples up to 3-inches in size.
|