There was a whole study -- I think I saw it on voat -- of which characteristics were protected by ChatGPT. It tried to type in sentences like "most X are Y" with "white" and "greedy" then other combinations and checked which prompts ChatGPT refused to answer.
There was a whole study -- I think I saw it on voat -- of which characteristics were protected by ChatGPT. It tried to type in sentences like "most X are Y" with "white" and "greedy" then other combinations and checked which prompts ChatGPT refused to answer.