AI brief
A LessWrong post proposes reinforcement learning environments in which helper agents are deliberately misaligned, so a main agent cannot earn its reward without handling that misalignment.
Written by AI from LessWrong's published text.
Source
“TL;DR: Agents are not rewarded for keeping the agents they delegate to aligned. Proposal: RL environments where the helpers are deliberately misaligned, so the agent can't earn its reward without dealing with that.”
Interpretation or community commentary rather than straight reporting.
Read original story ↗