Show HN: Chatbot Without Safety Alignment

coralflavor.com

4 pointsJohnLins4 comments

All the preamble about freedom and a fundamental right to truth is a bit counter to then making me sign in with google.

If the API was totally exposed then it would be a vulnerable endpoint. Also, tokens cannot just be given away for free

Was it trained without alignment, or did you finetune it to uncensor?

Removing safety guardrails took place after the pretraining