Comment by chii
1 day ago
I wonder if there's room in using AI to gather past edits of someone, as part of vetting, and use the sentiment analysis to check how neutral their biases are.
1 day ago
I wonder if there's room in using AI to gather past edits of someone, as part of vetting, and use the sentiment analysis to check how neutral their biases are.
AI is itself biased because the training data is.
Neutrality != necessarily accurate or useful. And the most neutral thing to say is nothing.
And most LLMs probably have Wikipedia as a significant part of their training corpus, so there is a big ouroboros issue too.