Re:GPTZero case study discovers it's only accurate on less t...
by ClumsyPilot on Sunday February 19, 2023 @11:55AM
(#34859299)
Or. ChatGPT will overtrain on it's own data and go to shit the way google search did
[ Parent
| on HN ]
Re:GPTZero case study discovers it's only accurate on less t...
by visarga on Sunday February 19, 2023 @12:44PM
(#34859869)
Training on its own data is a tradition already. For example RLHF example pairs rated by humans are generated by the model. So even our best models trained on their own outputs + rating from human labellers. The internet is a huge rating machine, AI will distill this signal and improve even while ingesting its own text.
Re:GPTZero case study discovers it's only accurate on less t...
by andyjohnson0 on Sunday February 19, 2023 @04:53PM
(#34862240)
Meta-ChatGPT's loss function optimises for ChatGPT generating training data that maximises the shittyness of Google's LLM.