when training a LLM would you prefer to first expose it to traumatizing content and then teach it to suppress the knowledge in its responses, or would you keep it innocent by removing the dark damaging topics from the training data?
if you've experienced trauma yourself, if you could erase the experience and all cascading damage/therapy from your memory & body, would you return to innocence?
if you have a child do you try to protect them from trauma?
More notes on nearby ideas
Quote tweet · Mar 14, 2024@tenobrus rerun my favorite...Quote tweet · Oct 23, 2024is bad code traumatic?Artificial intelligenceQuote tweet · Mar 19, 2023hah pretty close to 50/50 final result - I was really curious abt this one! https://x.com/mimi10v3/status/1637230393773391874Quote tweet · Dec 13, 2024thinking about this series of polls again https://x.com/mimi10v3/status/1637230393773391874