This lecture discusses photonic extreme learning machines (ELMs) and their relationship with reservoir computing. It begins with a comparison of multilayer perceptrons, ELMs, and reservoir computing, explaining the basic principles of each. The instructor elaborates on the architecture of multilayer perceptrons, highlighting the role of nonlinear activation functions and the backpropagation algorithm for training weights. The lecture then transitions to reservoir computing, emphasizing its fixed, high-dimensional connections and memory aspects, which allow for effective time series predictions. The advantages of using photonic systems for reservoir computing are presented, including high speed and low energy consumption. The instructor also covers programming techniques for photonic ELMs, showcasing how optical systems can be utilized for training without traditional digital methods. The lecture concludes with examples of applications in optical computing, demonstrating the potential of these systems in various tasks, including classification and regression, while addressing challenges such as overfitting and the importance of optimizing programming parameters.