Samuel Pedrielli

RLHF might be al­ign­ing the wrong thing. A differ­ent ap­proach.

Ego-Cen­tric Ar­chi­tec­ture for AGI Safety v2: Tech­ni­cal Core, Falsifi­able Pre­dic­tions, and a Min­i­mal Experiment

Ego‑Cen­tric Ar­chi­tec­ture for AGI Safety: Tech­ni­cal Core, Falsifi­able Pre­dic­tions, and a Min­i­mal Experiment

• Samuel Pedrielli 29 Jul 2025 21:54 UTC

  1 point

  0 ∶ 0

  on: Ego‑Cen­tric Ar­chi­tec­ture for AGI Safety: Tech­ni­cal Core, Falsifi­able Pre­dic­tions, and a Min­i­mal Experiment

  Thanks for reading — I welcome technical feedback on the three falsifiable predictions and the minimal experiment design. I’ll respond within 24–48h.