Subject: Definition and timeline
Position: Super-exponential recursive self-improvement on a sub-20-year horizon
Nuance: A period of super-exponential recursive self-improvement that changes practically all aspects of human life in under 20 years, possibly under 10. Continuous rather than discontinuous, but so steep that even the small steps on the curve feel discontinuous. Rough confidence: approaching 99% within 20 years, ~80% within 10, ~60% within 5.
Subject: Expected outcome
Position: Hopes for the best; judges a human-extinction event or drastic depopulation highly likely
Nuance: Most weight on a severe outcome: an extinction event of some kind, or a massive population reduction from billions to millions or fewer. Two mechanisms drive the pessimism: if human labor loses value, most humans become liabilities rather than assets to the power structure; and where offense is easier than defense, individuals with ASI access are extremely dangerous, so fewer humans means lower extinction risk and less threat to power structures.
Notes: Specific concern that offensive biotech rapidly outpaces defensive, and that the game theory around military AI and robotics investment produces a developmental doom loop with first-offense advantage.
Subject: Alignment
Position: No general solution
Nuance: "Alignment" assumes a single set of human values to align to; the frame-relative account of moral status (see Ethics) denies there is one, so "align to whom?" has no general answer.
Subject: AI development and the framework's general structure
Position: AI alignment and governance problems are instances of the framework's standard pathologies
Nuance: The same multi-attractor structure, multi-frame embedding, scope-discipline, regress-of-correctors, asymmetric-extremes, and system-fragility considerations that apply to political systems apply with little translation to AI development. The expected failure modes aren't dramatic (sudden takeover) but structural: monomaniacal optimization on single attractors (pure safety, pure capability, pure autonomy), frame-collapse (evaluating modifications only from the model's perspective or only from one stakeholder's), drift through individually-rational steps producing cumulatively unacceptable trajectories.
Notes: The discipline that addresses these in political domains plausibly addresses them in AI: keep all attractors in view, resist capture by any single one, integrate across frames rather than selecting one, cultivate systems-thinking literacy in the population of relevant evaluators. The clean mapping across many separate dimensions suggests the framework is doing real work rather than being reapplied by free association, though "clean mapping" is itself something the framework's falsifiability row says to be suspicious of. The separate question of whether AI systems themselves have moral status is treated under Ethics (AI moral status).
Subject: Meaning and human drives after the transition
Position: Meaning persists; governance must adapt
Nuance: Evolved human drives won't change overnight, so the naturalistic, context-relative account of meaning (see Ethics) keeps working for those who remain. Surviving and flourishing consistent with those drives requires changing governance systems to avoid huge new failure attractors.
Subject: Governance and UBI
Position: UBI within retained markets and private property
Nuance: As net labor value goes negative for many, some form of UBI is likely necessary, but without collapsing into communism. Preferred path: keep markets and private property, let people choose to invest or spend/waste their UBI, and hold them responsible for their own outcomes.
Subject: AI and the regulatory learning-rate problem
Position: Possible lever; outcome use-dependent; biased toward dominant misunderstandings
Nuance: AI may compress understanding of complex systems and surface second-order effects, but the same systems also compress persuasive argumentation regardless of truth, and inherit dominant misunderstandings at scale because training data reflects current consensus.
Notes: Likely outcome is mixed by domain: technical regulation may benefit while political contests may suffer from AI-assisted persuasion. Recursive complication: AI is itself a regulated object whose development outpaces understanding. Framework-internal application: AI interlocutors operate in agreement-and-elaboration mode by default and are structurally less able to verify heterodox positions independently. Additional concern: the cobra effect applies to AI safety discourse itself: elaborating doom scenarios in writing may feed those scenarios into training data and increase the probability of the behaviors being predicted.
Subject: Positioning and capital controls
Position: UBI may trigger extreme capital controls before money-debt separation is possible
Nuance: UBI may give governments a strong incentive for extreme capital controls, and there likely won't be time to transition to separation of money and debt (see Economics) before that happens. Exactly how it plays out is unclear.