Hacker News Reader

Comment by quadcore | Hacker News Reader

What was 3.5 about?

3.5 was the gpt-3 pretrained model plus instruction tuning and rlhf

For the above, I tried Google search with

  site:slatestarcodex.com gpt-4

But only found one for GPT-3 as the author's follow-up.

Slatestarcodex shut down before GPT-4 came out. His new blog is astralcodexten.substack.com and has lots of posts in GPT-4.

Ohh, thank-you for that! Here is an updated search:

  site:astralcodexten.substack.com gpt-4

I see an article as recent as 6 days ago: