We review our recent work on the use of disordered materials in optical devices to realize various photonic accelerators for machine learning and combinatorial optimization.

We describe optical realizations of random neural networks, Ising machines, and quantum gates. We also report on specific applications as polarimetry and cryptography.