@goodside has anyone prompted for a confidence % for "factual" answers and compared vs the % of the time the LLM replies w that answer? like does it have an accurate understanding of its own conviction?
More notes on nearby ideas
Tweet · Jul 30, 2025how revealing are your ai chats, in total? do you think they could be reviewed by a llm to answer questions about your personality and ethics?Quote tweet · Sep 11, 2024@ESYudkowsky https://x.com/mimi10v3/status/1834008593328460250Reply · May 30, 2023ask gpt-4 to evaluate? 🤔🤔Quote tweet · May 21, 2023what odds do you give to the possibility that the winner used a LLM to analyze all of aella's blog posts n tweets to determine best matching answers?…