How to AI (Almost) Anything
MIT's Media Arts and Sciences program offers this graduate course on applying modern deep learning and foundation models to real-world data beyond text, including vision, audio, sensor data, medical signals, music, art, smell, and taste. Lectures and readings cover core AI principles before moving into multimodal systems that combine language with other data types, such as connecting text to images or sensing to actuation. Students engage in discussions and complete a significant research component, working through the process of applying AI methods to a novel modality of their choosing. The course builds critical thinking about current technical achievements in AI and gives a firsthand look at how AI research is actually conducted. Materials are published through MIT OpenCourseWare under a Creative Commons license and include the syllabus, readings, and assignment structure.