Mark Poiler
@mpoilerfx • 1,486 subscribers
vibe poster
Videos
MIT put its full multivariable calculus course online in 2007 and charged nobody a dollar for it. Lecture 34 is the final review. A professor in an orange sweater writes Unit 1 on a chalkboard and starts working through the whole semester. The camera never moves much. No editing, no music, no production budget visible anywhere. MIT OpenCourseWare launched in 2001 with 50 courses. The plan was to publish materials from every course the university taught. Faculty were told they would get no royalties and no extra pay. Most agreed anyway. The catalog passed 2,400 courses. Downloads run past 300,000,000 and the site draws millions of visitors a year from countries with no equivalent institution. Universities charging 60,000 dollars a year watched a peer give the lectures away and kept charging. The thing people pay for was never the lecture.
Mark Poiler1,867,988 görüntüleme • 13 gün önce
The NSA kept 1 lecture in a vault for 40 years. In 2024 they released it. It's Grace Hopper explaining the future of computing in 1982. The channel that posted it: the National Security Agency's own YouTube. Hopper stands at a podium in Navy uniform, a captain, and talks to a room of NSA staff for 90 minutes across 2 reels. The title: "Future Possibilities: Data, Hardware, Software, and People." She compares the computer industry to the early automotive one, all custom parts and no standards, and argues software will become the hard problem, not hardware. She was right by decades. She invented the first compiler and gave COBOL its shape. She kept a clock that ran backwards on her wall to prove rules can be questioned. The tapes were too degraded for normal release, so the agency restored them before posting. 40 years in an NSA archive. 90 minutes. Now free.
Mark Poiler546,134 görüntüleme • 15 gün önce
A Princeton probabilist explains why enormous random matrices stop behaving randomly and start behaving like a single fixed object. Almost nobody watches it. This is Ramon van Handel at Harvard's Science Center, April 2025, on the strong convergence phenomenon. Every weight matrix in every model starts as random numbers. The claim here is that as dimensions grow, the extreme behavior of such matrices, the largest eigenvalues, the operator norm, stops fluctuating and locks onto a deterministic limit. That is why initialization works at all. Why spectral norms are predictable. Why large networks behave more consistently than small ones instead of less. Watch how he sets up what strong means. The distinction between convergence of averages and convergence of the extremes carries the entire result. A machine learning engineer I know rewatched the setup twice and said scaling laws stopped feeling like empirical luck. Free on YouTube from the Harvard Mathematics Department, subtitles on. Random at small scale. Deterministic at large. Nobody told the engineers.
Mark Poiler102,004 görüntüleme • 7 gün önce
Daha fazla içerik yok.