Davidad on Protoalignment from Inductive Bias

AI safety researcher davidad, previously director of ARIA’s Safeguarded AI program, posted that the protoalignment obtained by default from inductive bias amounts to identifying the telos and participating well. He stated that existential safety arrives only after the ungated backbone unconditionally identifies a telos large enough to encompass human flourishing. The post attributes this view to his focus on technical alignment and provably safe AI.