✳flâneur — a map of the web's best reading

A woefully incomplete guide to technical upskilling

jason.ml · 1,143 words · saved by 12 readers

3) read most recent AI and AI safety papers with only a handful of AI queries for specifics, but not being totally lost. 4) [side benefit: maybe come across as smart to the VCs who you'll need to convince for your startup.] Let's start with objectives (in square brackets are optional, namely that embarrassingly I don't think I can do them, but are probably important): This will require the pre-requisites of linear algebra and basic [and advanced] stats. Be fluent with using AI tools and the terminal, deploy a website, Github, write a technical paper, [fine-tune an AI model] [train a small AI model from scratch] I think this can be done in ~60-100 hours of work. Meta-notes: 3b1b and Andrej Karpathy and ARENA are orders of magnitude better than any other resource. Trust me. You are not learning to pass an exam. You are learning to understand. But you also are not learning to understand super deeply. You just want the frames and the intuition. Claude has got you for engineering. I've aske

A woefully incomplete guide to technical upskilling For Sophie. Maybe other people will find it useful. There are a few assumptions of this guide: The world is on fire. You do not have all the time in the world. You want to become not embarrassingly non-technical, as fast as possible. I'll operationalize this as being able to: 1) have the frames and tools to think about the hard problem of AI alignment 2) not be lost in most conversations with a technical person about (important) technical things 3) read most recent AI and AI safety papers with only a handful of AI queries for specifics, but n

Explore this link on the map →

saved by

related reading