News for hackers. Stuff that matters.

A comment on Anthropic publishes the 'system prompts' that make Claude tick. Parent comment

Re:Anthropic publishes the 'system prompts' that make Claude... by cjbillington on Tuesday August 27, 2024 @03:49PM (#41372149)
What do they do instead? Given we're not talking to a base model.
[ Parent | on HN ]
Re:Anthropic publishes the 'system prompts' that make Claude... by tqi on Tuesday August 27, 2024 @04:44PM (#41372788)
Supposedly they use "RLAIF", but honestly given that the first step is to "generate responses... using a helpful-only AI assistant" it kinda sounds like RLHF with more steps.

https://www.anthropic.com/research/constitutional-ai-harmles...

[ Parent | on HN ]