GPT 4

1. GPT-4 finished training four years ago today (twitter.com via hn)

   +41 6w gpt-4

   GPT-4 finished training four years ago today. - To celebrate its birthday, how about telling us how many parameters it had?

2. Weak-to-strong generalization (GPT-4 Turbo was Pareto-SOTA) (www.forourposterity.com via hn)

   +1 6w gpt-4

   Weak-to-strong generalization A new research direction for superalignment: can we leverage the generalization properties of deep learning to control strong models with weak supervisors? I'm very excited and proud to share some of the work…

← all threads