alcinnz, Datasette 1.0a4 and 1.0a5, plus weeknotes - Simon Willison: http://simonwillison.net/2023/Aug/30/datasette-plus-weeknotes/#atom-everything
Llama 2 is about as factually accurate as GPT-4 for summaries and is 30X cheaper -Waleed Kadous @ anyscale:https://www.anyscale.com/blog/llama-2-is-about-as-factually-accurate-as-gpt-4-for-summaries-and-is-30x-cheaper
Simon Willison's excitement: https://simonwillison.net/2023/Aug/30/llama-2-summaries/
Biased, but I have long strongly suspected that larger models aren't really better... That there's better innovations to take from this hype cycle...
Add comment