Post

Anil Dash
@anildash.com

Can you imagine talking about another human being with this level of dehumanization and thinking it’s funny? Like, imagine saying this in a room in front of a stranger and supposing that they would not comment when they heard you. And I’m expected to act like it’s normal, or *I’m* weird to respond.

This post is unavailable.
asura
@asura.dev

Now imagine if you can use Olmotrace tools on the LLM you've made. What you've done is not a bad idea at all and with enough tooling you can create some pretty interesting introspection. I can't remember what I did yesterday, so trying to remember what I wrote 25 years ago would be impossible.

allenai.orgGoing beyond open data – increasing transparency and trust in language models with OLMoTrace | Ai2OLMoTrace lets you trace the outputs of language models back to their full, multi-trillion-token training data in real time.
asura
@asura.dev

Every day that goes by I really think there was a psyop against LLMs just like there's that funky psyop against Wikipedia Being able to do what you've done paired with something like olmotrace and also being able to use the classic Feynman Technique against an LLM would actually be uh very good