The next milestone in experimental quantum information processing is the demonstration of a quantum advantage, i.e. the unambiguous demonstration of an information processing task, no matter how trivial, at which a quantum device outperforms a classical computer. In photonics, this can be done by sending single photons through a linear network, a problem known as boson sampling. A major practical hurdle on the way to constructing a boson sampler is that it is poorly understood how imperfections in the system, such as photon loss and finite distinguishability (wave function overlap) degrade the hardness of simulating the quantum interference occurring in such a device.

In this talk, I will address this problem by demonstrating a classical algorithm which uses these two imperfections to perform an efficient classical simulation of a boson sampler. For situations where this algorithm works (i.e. if there is enough loss / imperfections), we can rule out a quantum advantage. This algorithm can therefore be used to guide the design of future photonic quantum advantage demonstrations. I will show that strong qualitative improvements to photonics technology are needed to demonstrate a quantum advantage using photonics.