Yannic Kilcher occupies a specific and valuable niche, which is helping people who already work in or study machine learning actually understand the research rather than just skim the headlines about it. His paper review videos take a single piece of work, often a landmark result or a very recent release, and walk through it section by section, explaining the core idea, the method, the results and, crucially, whether the claims hold up. What sets him apart from a lot of AI content is the critical stance, because he is comfortable questioning hype, flagging weak evaluation, and pointing out when a much celebrated paper is doing less than it appears to, and in a field awash with breathless announcements that scepticism is genuinely useful. Watching enough of these does something subtle and important, it teaches you how to read a paper yourself, because you absorb the questions he asks and the things he checks for, and that is a skill that outlasts any single explanation.
Alongside the deep dives there are news roundups and interviews that give a good feel for the pace and the personalities of the field. The honest framing is that this is not a course and never pretends to be. There is no curriculum, no ordering, no exercises and no assessment, and the topics follow his own curiosity rather than any structured progression, so it works as a companion to your learning rather than the spine of it. It also assumes a real background, because he does not stop to explain what a gradient or an attention head is, and a newcomer will quickly feel lost.
The videos are dense and move quickly, so getting the full value means pausing, rewinding and sometimes reading the paper alongside him. My honest opinion is that for a practitioner or a serious student, this is close to essential, both as a way to stay current without drowning and as a masterclass in reading research critically. Just come to it once you have the foundations, and treat it as the thing that keeps you sharp rather than the thing that gets you started.