Comment by throwawayk7h
3 days ago
the alarming thing to me is that the prompt tweak provided should not have caused the model to start spewing pro-nazi nonsense.
3 days ago
the alarming thing to me is that the prompt tweak provided should not have caused the model to start spewing pro-nazi nonsense.
Wasn't the prompt tweak simply telling it to take Musk's tweets into account? If anything, the result was entirely predictable.