4 ms·
My well-documented melancholy around the state of the LLM “conversation” notwithstanding, I’ll point out that there’s a long and generally productive history of
by benreesman 3y ago
My well-documented melancholy around the state of the LLM “conversation” notwithstanding, I’ll point out that there’s a long and generally productive history of adversarial training: from the earliest mugshot GANs to AlphaZero, getting these things to play against each other seems to produce interesting results.
Whatever the merits of this or that “ChatGPT detector”, the concept isn’t unprecedented or ridiculous.