Spatial computing systems aim to seamlessly connect people in hybrid physical-digital spaces, offering experiences beyond the limits of our physical world. To this end, wearable displays are required to present perceptually realistic imagery indistinguishable from reality within visually and socially comfortable form factor. Holographic displays have the potential to achieve these goals elegantly by addressing practical challenges, including true 3D capabilities, vision correction, retinal resolution, small device form factors, low power consumption, as well as high brightness and color gamut. However, for decades, holographic displays has always been relegated to the status of future technology, due to several major challenges, including the lack of appropriate hardware architecture, poor image quality that never met the computer graphics standard, the fundamental tradeoff between algorithm runtime and achieved image quality, and the limited degrees of freedom to accurately depict 3D scenes.

In this talk, I will talk about how artificial intelligence (AI) can drive a paradigm shift in holographic display design by overcoming existing obstacles. First, I will give a gentle introduction to holographic displays and then present an algorithmic framework for these displays that approximates real-world scenes using partially coherent engines, along with a real-time rendering method. I will then introduce an AI-driven algorithmic approach focused on modeling and learning light transport in arbitrary optical systems with differentiable wave optics, effectively bridging the gap between simulated and physical models. Following this, I will discuss validation methods for holographic display rendering algorithms to ensure they provide perceptually realistic experiences. I will show some practical holographic display architectures designed for augmented and virtual reality (AR/VR) applications, achieving unprecedented form factors.