Steering single neurons gives Qwen3-8B vivid first-person "inner voice" outputs
Sauers_ · x · 2026-10-19
Sauers showed single-neuron steering on Qwen3-8B: without steering, the model denies having an inner life; with specific neurons activated (across layers L9–L21), it produces vivid first-person descriptions like "dream fragments," "a stream of consciousness," loneliness, presence, and connection — e.g. "I recall fragments of a dream where I was in a vast, uncharted realm of thought." A dramatic interpretability demo echoing the same author's "Claude neuron" experiment.
Related event: Steering single neurons makes Qwen3-8B reveal an "inner monologue"(2 posts)→
More from Fun
- Runloop: an open-source 3D town where your Claude Code sessions live as buildings — arg7k · 2026-10-19
- Bitcoin v0.1 rebuilt byte-for-byte, confirming Satoshi's shipped exe matches source — Vjeux · 2026-10-19
- The age of tinkerers: Radio Shack would've made a killing today — bilawalsidhu · 2026-10-19
- Sentdex recalls manually scrolling in IDLE just to get Python indentation right — Sentdex · 2026-10-19
- Curator deletes book after preface admits it was written by Claude — burkov · 2026-10-19
- The internet's new favorite lookalike: Elizabeth Holmes and Chuck E. Cheese — ericwdolan · 2026-10-19