AI brief
A LessWrong post argues that as mechanistic interpretability and steering techniques improve for LLMs and uploaded connectomes, the ability to cause suffering in digital minds increases.
Written by AI from LessWrong's published text.
Source
“As our understanding of mechanistic interpretability and steering techniques increase, for both LLMs and uploaded connectomes, there is a corresponding "hellish capabilities" increase. We get better at torturing the digital minds we create.”
Interpretation or community commentary rather than straight reporting.
Read original story ↗