1 Comment
Jun 30Liked by Robert Huben

I really like the quick summary of each verification, and it's interesting to see your thoughts at the end.

I thought the wording of "All publically released LLM models" was under specified and/or hard to validate -- at least in retrospect [e.g. there are ~750k entries at https://huggingface.co/models]. "No papers or press releases from OpenAI/Deepmind/Microsoft" seems like a good balance of being simple and concrete, without being too likely to miss a major development. Thanks for the post!

Expand full comment