After 30 years of R&D, photonic integrated circuits (PICs) are finally ready to serve the general consumer marketplace. Artificial Intelligence (AI) will provide the first entry vector, for 3 reasons: 1. Power Consumption - The extreme market pressure to move AI to the network edge places serious demands on the power efficiency of computational hardware. Thanks to fundamental physical differences, as well as certain power amortization characteristics, photonic computing engines have an inherent 1,000x advantage over their electronic counterparts. 2. Fault Tolerance - Current generation photonic computing architectures use low precision analog techniques for performing matrix-vector multiplication, which limits their general applicability. However, they are a perfect fit for AI algorithms, which are not only tolerant of computational errors, but often introduce them purposefully, for increased robustness. 3. Computational Requirements - The raw computational horsepower needed to support near-term future AI needs is immense. And photonics alone is equipped to fulfill those needs at an acceptable rate of power consumption.
David began his career in photonics nearly 30 years ago in 1990, when he entered graduate school at the University of Colorado at Boulder, working in the Optical Computing Systems Center. After graduating in 1994, he went to work for one of the center's spin-off companies; Displaytech, Inc., where he performed research on novel "smart pixel" liquid crystal on silicon (LCOS) devices intended for use in optical information and signal processing systems. One such device: the universal spatial light modulator (USLM) resulted in a Best Paper award from the Society of Information Display. David now serves as Chief Architect at Luminous Computing, Inc., a new start-up in the photonics AI acceleration space.