Research
96d agoNew Theory Challenges Goal-Oriented AI Alignment
A new essay in The Gradient argues that rational agents should lack fixed goals. Instead, the author proposes aligning AI to practices—networks of action-dispositions and evaluation criteria. This shifts the focus from the orthogonality thesis toward virtue ethics.
The Signal
Practitioners must now consider if goal-directed architectures fundamentally misrepresent how human rationality actually functions.