Hacker News Reader

I'm fining letting the models decide. In fact, they will only get more nuanced and sophisticated over time. This is the worst it will ever be.

I'm using Sonnet right now, I'd be curious to see how Opus would classify the same claims.

It's not an AI issue, there's human based fact checks that similarly "confirm" as "true" the non-interesting bit of a claim.

Snopes has a "Correct attribution" label for this, which also covers ambiguous jokes, misstatements etc.

hey i like this, i'm goinna adopt the snopes scale