MIT 6.S191

MIT 6.S191 is taught by Alexander Amini and Ava Soleimani and has run every year since 2017, with lectures re-recorded annually so the content stays current recent editions cover transformers, diffusion models, and large language models alongside the classical foundations (CNNs, RNNs, deep generative models, reinforcement learning). Each lecture is roughly 45 minutes to an hour, paired with publicly available slides and lab notebooks (in PyTorch/TensorFlow) that let you implement what you just watched. It's more compressed than a full-semester course like CS231n or NYU's Deep Learning course, so it works well either as a fast, current overview if you already have background, or as a way to sanity-check that your existing knowledge is up to date with 2025/2026-era techniques. Because MIT re-does it every year, it's one of the few free resources on this list that won't feel dated in the way some 2017-era university lectures now do.