AI brief
A post on LessWrong discusses whether a model's misalignment can be predicted from its training data before training, since fine-tuning on subtly flawed data can cause broad misalignment that is currently caught mostly after training through auditing.
Written by AI from LessWrong's published text.
Source
“Fine-tuning on subtly flawed data can make a model broadly misaligned. Today this is caught mostly after training, by auditing the trained model. We ask whether it can be predicted beforehand, from the training data.”
Interpretation or community commentary rather than straight reporting.
Read original story ↗