[zeryx]

It's as good as an intern today, there's definitely room to improve to avoid saying incorrect things - great work being done on providing negative samples during RLHF training; also some papers working on incorporating trust (not near my laptop so don't have links)

Textbooks are all you need is another extremely interesting area of focus.

Improved correctness and a general reduction in hallucinations I predict will end up hitting OSS models by EOY

[chongli]

How is that "as good as an intern?" Interns, for the most part, do not hallucinate and confidently lie on topics they know nothing about. Good interns tend to ask a lot of questions and quickly learn the cultural cues around the office, allowing them to fit in and become productive. LLMs are incapable of any of this.

John Schulman from OpenAI recently gave a talk about the hallucination and uncertainty issues [1]. He claimed that models already have information about their uncertainty but that it remains an open problem as to how to express that uncertainty in natural language. One of the big issues with trying to prevent the model from hallucinating is that you can err too much on the side of caution, causing the model to lie about things it actually does know the correct answer to.

[1] https://www.youtube.com/watch?v=hhiLw5Q_UFg

[pixl97]

>do not hallucinate and confidently lie on topics they know nothing about.

Oh shit, we've developed digital salespeople...