Conversational AgentSemantic Search
This site has been created by McGill students Henri Lemoine, Fraser Lee, and Thomas Lemoine as an attempt to create a "conversational FAQ" that can answer questions about AI alignment. When asked a question, we
- Embed the question into a low dimensional semantic space
- Pull the semantically closest passages out of a massive alignment dataset
- Instruct an LLM to construct a response while citing these passages
- Display this response in conversational flow with inline citations
We've created this as an attempt on the $5k bounty for a LW conversational agent.
We hope that it can be a useful tool for the community, and we're eager to hear feedback and suggestions. For a technical report on our implemention, see our LessWrong post.
Since this is still an early test, all questions and answers are stored.
Zero other information is collected.