John, would you think, based on your background that one could develop a series of questions to feed an AI source that would basically reveal its calibration or, even ideally, to force calibrate it.
-- questions about the most biased genre's that push it to talk on those areas and reveal bias?
John, would you think, based on your background that one could develop a series of questions to feed an AI source that would basically reveal its calibration or, even ideally, to force calibrate it.
-- questions about the most biased genre's that push it to talk on those areas and reveal bias?
-- questions that feed it additional info followed by questions to determine if it utilized that info vs before feeding it?
-- feed specific info to different AI sources before asking the same question in multiple ways; analyze responses to compare the different AI's
-- some combo of above to get ALL AI sources(say, compare 2 or 3) to vomit the SAME response.
John, would you think, based on your background that one could develop a series of questions to feed an AI source that would basically reveal its calibration or, even ideally, to force calibrate it.
-- questions about the most biased genre's that push it to talk on those areas and reveal bias?
-- questions that feed it additional info followed by questions to determine if it utilized that info vs before feeding it?
-- feed specific info to different AI sources before asking the same question in multiple ways; analyze responses to compare the different AI's
-- some combo of above to get ALL AI sources(say, compare 2 or 3) to vomit the SAME response.
Yes, probably best to test AI bots against each other. See what results they favor. Same with search engines.