AI brief
A LessWrong post argues that fixed-weight models will always be vulnerable to adversarial examples in their concept spaces and will therefore be misaligned under sufficient optimisation pressure.
Written by AI from LessWrong's published text.
Source
“This post argues that fixed-weight models (at least as we understand them today) will a) always be vulnerable to adversarial examples in their concept-spaces, and b) hence will be misaligned, under sufficient optimisation pressure .”
Interpretation or community commentary rather than straight reporting.
Read original story ↗