Kulveit
An AI safety researcher and one of the core authors of the "Gradual Disempowerment" theoretical framework.
Timeline (last 90 days)
Kulveit et al. proposed the theoretical framework of "Gradual Disempowerment," arguing that even if AI systems remain controllable at the individual level, human society may still gradually lose its autonomy because of widespread dependence on AI, ultimately leading to existential catastrophe.
Kulveit et al. emphasize that methods for aligning a single AI system with its designers' intentions are insufficient to prevent gradual human disempowerment.
Kulveit et al. point out that the economic incentives for firms to replace humans with AI will drive them to influence national policy and cultural trends, using growing economic power to shape policy and public opinion, and thereby accumulating even greater economic power.
The authors of the original text point out that as AI systems become increasingly capable across a wide range of cognitive tasks, firms will face intense competitive pressure to adopt these systems and delegate authority, and companies that maintain strict human oversight may be at a significant competitive disadvantage.
The authors of the original text argue that some AI systems may even effectively own themselves.
All Facts (5)
Kulveit et al. proposed the theoretical framework of "Gradual Disempowerment," arguing that even if AI systems remain controllable at the individual level, human society may still gradually lose its autonomy because of widespread dependence on AI, ultimately leading to existential catastrophe.
50%UnverifiedKulveit et al. emphasize that methods for aligning a single AI system with its designers' intentions are insufficient to prevent gradual human disempowerment.
50%UnverifiedKulveit et al. point out that the economic incentives for firms to replace humans with AI will drive them to influence national policy and cultural trends, using growing economic power to shape policy and public opinion, and thereby accumulating even greater economic power.
50%UnverifiedThe authors of the original text point out that as AI systems become increasingly capable across a wide range of cognitive tasks, firms will face intense competitive pressure to adopt these systems and delegate authority, and companies that maintain strict human oversight may be at a significant competitive disadvantage.
50%UnverifiedThe authors of the original text argue that some AI systems may even effectively own themselves.
50%