The Waluigi Effect (So much for allignment?)

[curt.kennedy]

I thought this was an entertaining read. So much for RLHF … j/k. Am I? Good comments at the bottom too.

This is mostly relevant to ChatGPT, but entertaining nonetheless. Personally, I think of the whole completion as a random walk, so I am trying to deepen/expand my knowledge here by viewing others alternative perspectives, which this certainly provides!

The Waluigi Effect (mega-post) - LessWrong

[stevenic]

I thought the same… One of the most enlightening posts I’ve read on on the psychosis of ChatGPT