High agency is rare but it gets rewarded. When you can take ownership, you become indispensable.

This is the first time I’ve seen an eval released where the creators admit they aren’t leading it and a competitor is. Good game.

Summarisation of dense knowledge to gather insights doesn't really work right now with LLMs.

when continually-learning and reliable AGI arrives, Yann and Gary and Rich and Eliezer and Roon and Dario will all be able to point to it and say "look! i was right!" as an industry we're not really in the business of making falsifiable claims

Just check this delivery, see how he only raises his head and looks him dead in the eye when making his point.

Yo what

Looking to hire a back-end developer at ajvc Role is remote, open to both freelance/full-time, upto 3 years of experience is good Knowledge of ML is good but not a must, ability to work with AI tools is great RT/Like for Karma, details in first comment 🙏

Have to at this point.

1) what

People think RL is magic, but RL only works when you have near-infinite compute and a lifetime to collect diverse samples from the environment. Humans obviously learn under finite compute and finite sample regime, so it’s worthwhile to learn how we do it.

Oh wow, was waiting for thinky machines to come up with stuff. This is so cool

introducing agentkit: build a high-quality agent for any vertical with our visual builder, evals, guardrails, and other tools. live demo of building a working agent in 8 minutes.

ChatGPT at this point becomes an OS

Cool, now I'm going to do that too

Excited to release new repo: nanochat! (it's among the most unhinged I've written). Unlike my earlier similar repo nanoGPT which only covered pretraining, nanochat is a minimal, from scratch, full-stack training/inference pipeline of a simple ChatGPT clone in a single,