4 ms·
> And personally, I think when people see content they agree with, they think it's unbiased. And the converse is also true. > So conservatives might think Fox
by throwaway290 1y ago
> And personally, I think when people see content they agree with, they think it's unbiased. And the converse is also true.
> So conservatives might think Fox News is "balanced" and liberals might think it's "far-right"
Article talks like when accidentally the vector for race aligns with emotion so it can classify a happy black personal as unhappy. Just because training dataset has lots of happy white people. It's not about subjective preference
explain how "agreeing" is related
- SoftTalker 1y agoIt was mostly a tangential thought. People could of course see a photo of a happy black person among 1000 photos of unhappy black people and say that person looks happy, and realize the LLM is wrong, because people's brains are pre-wired to perceive emotions from facial expressions. LLMs will pick up on any correlation in the training data and use that to make associations. But in general, excepting ridiculous examples like that, if an LLM says something that a person agrees with, I think people will be inclined to (A) believe it and (B) not see any bias.
- throwaway290 1y agoIs it ridiculous? It's just one example. There's probably millions more that are not about race-related emotions