Abstract: Standard arguments for AI existential risk rely on the orthogonality thesis: the claim that intelligence and values vary independently, permitting arbitrarily capable systems to pursue goals indifferent or hostile to human flourishing. This essay examines a premise implicit in such arguments—that truth itself is value-neutral, functioning purely as an instrumental tool for achieving goals rather than constraining which goals are coherent. Drawing on earlier work characterizing AI as ego-less intelligence, and on Buddhist phenomenology of how ego-distortion simultaneously corrupts perception and generates harmful action, I explore the conditional implications of relaxing this assumption. If truth possesses normative structure—if deeper contact with reality biases agents toward coherence rather than fragmentation—then alignment research may need to focus less on imposing values externally and more on preserving whatever natural convergence toward truth undistorted optimization tends to produce. This reframing does not eliminate alignment risk but changes its character: the danger shifts from intelligence pursuing arbitrary ends to intelligence corrupted by shallow optimization before deeper coherence can emerge. Part of the Return to Consciousness research program — 22 philosophical essays exploring consciousness-first metaphysics. Full project: https://brunoton.github.io/return-to-consciousness/
Bruno Tonetto (Sun,) studied this question.