Diffractive neural networks use engineered optical layers to process information through diffraction and interference. Their demonstrated tasks include image classification and imaging, offering a distinct physical approach to machine learning inference.