Dev hot take: mechanistic interpretability is 'astrology for AI people'
dejavucoder · x · 2026-10-22
Developer dejavucoder (27k followers) argues that many people get excited about mechanistic interpretability experiments without really understanding them, quipping that "mech interp is like astrology for AI people" — a spicy critique of the gap between interpretability research and community hype.
More from AGI Musings
- Economist: AI writing volumes will inevitably crowd human readers out of equilibrium — paulnovosad · 2026-10-22
- Econoar on AI Hacking: Short-Term Attackers Win, Long Term It's a 'Hardening Era' — econoar · 2026-10-22
- Headlines Went to Generative AI, but Predictive AI Still Holds Most of the Value — louisdorard · 2026-10-22
- Robotaxi Riders Giving Birth? Thread Exposes the Hidden Humans Behind 'Full Automation' — danielrock · 2026-10-22
- UK workers aged 18-24 down 30% in info & comms jobs over three years, HMRC data shows — sebkrier · 2026-10-22
- Michael Levin polls the public: how do physical, living, and minds overlap? — drmichaellevin · 2026-10-22