Engineer to AI researchers: transformers are soft automata, not minds
gerardsans · x · 2026-09-19
A software engineer argues that transformers are just "soft automata": attention latches to context states and produces transitions — a program, not a self or a mind. He contends that priors shaping outputs is exactly the point of instruction-following RL post-training, and blames the anthropomorphic "alchemy" in AI discourse on practitioners lacking software engineering backgrounds, while labs no longer employ engineers to connect the dots. A pointed contribution to the ongoing debate over whether LLMs exhibit genuine cognition.
Related event: Engineers Argue LLMs Are Just State Machines, Not Minds(2 posts)→
More from AGI Musings
- AI community clashes over 'rogue agents' narrative: Redwood and METR quotes fuel the debate — Dan_Jeffries1 · 2026-09-19
- Fields Medalist Villani 'shaken' by OpenAI's claimed millennium problem solution — akbirthko · 2026-09-19
- Most Jobs Don't Involve Einstein-Level Science — and Breakthroughs May Not Boost Growth — binarybits · 2026-09-19
- Thom Wolf: big labs are air-gapped, but researcher departures still leak information — Thom_Wolf · 2026-09-19
- Timothy B. Lee: his wait-and-see AI policy argument fails under 'fast takeoff' — binarybits · 2026-09-19
- 'Doomers are not serious people': the GPT-6 release reignites AI risk debate — inductionheads · 2026-09-19