factual

Models use multiple parallel circuits, replacing shortcuts with deeper reasoning

Circuits work shows there is never a single path for why a model does something but always multiple paths of varying depth; for refusing a harmful 'bomb' request, there is a direct shortcut path and a separate deeper reasoning path, and one narrative is that as the model gets smarter during training it learns to replace shortcut imitation with deeper reasoning circuits while retaining the harmless shortcuts.

factualpending

Speaker

Trenton Bricken

Evidence Quote

“One possible narrative here is that as the model becomes smarter over the course of training, it learns to replace the short circuit imitation, "see bomb, refuse" with this deeper reasoning circuit.”

Source

Is RL + LLMs enough for AGI? — Sholto Douglas & Trenton Bricken— Dwarkesh Patel Podcast
Created: 6/13/2026, 3:28:51 AM

Your notes

Loading notes...