Agree on raw language model outputs being unverified and untrustworthy. It’s a fine straw man to go after, but the vast majority of language model use cases for news (from the LA Times thing to the prediction use case in your piece) will have fine-tuning for specifics with far more control and value

Also on curation, isn't providing a different form of curation (including using AI data) a great thing for journalism to do to differentiate from X and other platforms built around engagement at all costs? In the ol' bundled newspaper days a physical daily provided many things to many people.

I don't know a lot about this particular "bias meter," but isn't having data about what a particular news article is doing (even if from AI) helpful in terms of generating user trust? I think the concept of the reader and their needs is important to include in these conversations.