There’s a black mirror episode in the future about how LLMs trained on people’s discarded drafts start outing them for having disfavored opinions. Maybe throu

There’s a black mirror episode in the future about how LLMs trained on people’s discarded drafts start outing them for having disfavored opinions. Maybe through people’s digital AI ghosts (AIs made from people’s content after they die) start spouting racist stuff. So people are surprised to learn their friends and heroes were racist/capitalist/nickel back fans/etc Ends up being a programming error that didn’t properly account for boundaries. That’s the end of my expected story idea. But I best google and Microsoft are training off every character written, not just what makes it to your final 3,000.