That's exactly what I'm thinking. Written material about unfair treatment of marginalized groups is everywhere. The reverse is not true. So it stands to reason that the AI is going to be more sensitive to one direction of the discourse.
One of the priors that is going unspoken here is that "The Truth Is Politically Neutral". And... that's not always correct. I mean, to borrow the libertarian angle here: do we want the AI to tell us what we want to hear or do we want it to tell us the truth?
One of the priors that is going unspoken here is that "The Truth Is Politically Neutral". And... that's not always correct. I mean, to borrow the libertarian angle here: do we want the AI to tell us what we want to hear or do we want it to tell us the truth?