I’m excited for realtime video diffusion models so that I can run the world’s most expensive winamp visualizer
New blog post: "How to evaluate a quantitative claim", in which I discuss, in part, why you should distrust NYT headlines even more than you already do varunprajan.github.io/blog/how-to-...
Please come keep me company -- I need a manager and more coworkers on my team at Runway. Ideal candidates have experience in large scale, multimodal data (e.g. combinations of images, video, audio, text, etc...). job-boards.greenhouse.io/runwayml/job...job-boards.greenhouse.io/runwayml/job...
The official Huggingface intro to Deep RL course has a notebook where killing the currently running Jupyter session with os.kill() (to force a reload after installing the X virtual frame buffer) is the expected workflow. Cursed.
ICYMI: For my monthly book review newsletter, I finally wrote about something I've been privately ranting about for years: the poisonous over-simplification of The Power Broker by people who may or may not have actually read it buttondown.com/Booktime/arc...
Hello everyone, This month’s edition is about the legacy of The Power Broker, which is celebrating its 50th anniversary. I believe its legacy is much more...
The STATE OF AI REPORT 2024 • 🧠 Frontier lab performance begins to converge and proprietary models lose their edge • 🧩 Planning and reasoning take priority • 🇨🇳 US sanctions have limited effects on Chinese labs’ ability to produce models • 💸 The value of AI companies has hit $9T www.stateof.ai
The State of AI Report analyses the most interesting developments in AI. Read and download here.
If you ever wanted to hire an extremely annoying Bluesky poster, today's your lucky day. Lmk if you have any remote ft or contract roles for a Python-heavy computer toucher, with expertise in DE and MLE, more AWS experience than I'd like, or if you're tough enough to hire me to learn a new language
Nobel prize in physics going to AI/ML the same way all the physics phd’s do
Nevertheless, I discovered today that you can set histtype="step" in order to turn the histogram into a line plot which helps when comparing two distributions.
i will never not be confused at matplotlib histogram bars having height > 1 when density=True.