Probabilistic computing with quantum light

The worlds demand for computing power is ever increasing. A prime example of this is the computationally very intensive artificial intelligence, which has found many applications ranging from speech recognition to autonomous driving. However, the traditional approach of building ever smaller transistors the building blocks at the very heart of computer chips to increase computing power is reaching its limits. Especially, for such computationally intensive systems there is currently a paradigm shift taking place, with researchers considering moving away from traditional computing components to specialized hardware that performs better in key metrics such as power efficiency. This is critical as the share of electricity used for computation is rapidly growing worldwide. A particularly promising research direction is to harness the intrinsic ability of light to perform computational tasks on an ultra-fast time-scale. This is possible because light is inherently fast and the way it moves through optical systems (think of for example an ordinary lens) mimics basic mathematical operations used in computing. In our research, we propose to demonstrate an optical computer based on light pulses. These light pulses will serve as a true random number generator, since the randomness will be based on a quantum effect. Equipped with these random numbers we can use an emerging computing paradigm which is called probabilistic computing that lies between classic (deterministic) and quantum computing. While it shares many of the advantages of quantum computing, it is not limited, by practical and engineering problems which plague quantum computing. In this way we plan to demonstrate for the first time that this particular approach can greatly increase our capability to tackle important computational challenges. The design of more efficient optical hardware for energy-hungry algorithms in AI is also relevant given the current challenges in reducing energy consumption worldwide.