View on Amazon Study modeFoundationsExpand all01What human-history growth modes preceded the AI-driven intelligence transition, and at what doubling times?02Why is the AI-driven intelligence transition different in kind from prior human growth-mode transitions?03What three properties make superintelligence a unique strategic problem?04State the basic argument of Superintelligence in three steps.05Name Bostrom's four x-risk categories.06Why does superintelligence cut across all four x-risk categories?Paths to SuperintelligenceExpand all01Name the five paths to superintelligence.02Why does Bostrom give de novo AI the highest probability?03What does the WBE path require, end to end?04What's the strategic difference of WBE-first superintelligence vs. de novo AI?05Name the three sub-paths of biological cognition enhancement, and which is the credible path to substantially superhuman cognition.06Why does Bostrom rank BCIs and human networks as the weakest paths?Forms and KineticsExpand all01Name the three forms of superintelligence.02Which of Bostrom's three forms of superintelligence is hardest to achieve and most strategically dangerous?03Bostrom models AI takeoff speed with two variables. Define them.04Why does Bostrom expect his ratio Π/R\Pi/RΠ/R (optimization power applied to a system over its recalcitrance) to favor fast takeoff once human-level AI is reached?05Define slow, moderate, and fast takeoff.06Why does fast takeoff change the entire strategic problem, not just compress its timeline?07What is hardware overhang, and why does it make fast takeoff more plausible?Decisive Strategic AdvantageExpand all01What conditions make first-mover advantage decisive?02What is a singleton?03Why does Bostrom argue takeoff dynamics tend to produce a singleton?04Name Bostrom's six cognitive superpowers.05What is the wise singleton thesis?06What's the strongest argument that a multipolar outcome is more realistic?The Superintelligent WillExpand all01State the orthogonality thesis.02What's the strongest argument against orthogonality?03State the instrumental convergence thesis.04Name the five convergent instrumental goals.05Of Bostrom's five convergent instrumental goals, which is most underestimated, and why?06Walk the paperclip-maximizer thought experiment. Why is the silly framing the point?Failure ModesExpand all01Define perverse instantiation with an example.02Why does perverse instantiation generalize beyond cute thought experiments?03What is infrastructure profusion?04What is mind crime?05Why is mind crime uniquely hard to detect or prevent?06Define the treacherous turn and its incentive structure.The Control ProblemExpand all01Name the four capability control methods.02Why does Bostrom regard all capability control methods as temporary?03Name the four motivation selection methods.04What's the central trade-off across motivation selection methods?05Distinguish oracles, genies, sovereigns, and tools.06Why isn't "tool AI" the easy answer to the control problem?Acquiring ValuesExpand all01State the value-loading problem.02Why is even "make humans happy" hard to specify?03What's the value learning approach?04What's the central technical challenge of value learning?05What is evolutionary selection as a value-loading method, and what does it cost?06What is motivational scaffolding, and what does it cost?07What is emulation modulation as a value-loading method, and how does its risk profile differ from de novo AI value loading?08How do multipolar scenarios offer an alternative to value-loading a singleton, and what's the hidden cost?Choosing Criteria and the Strategic PictureExpand all01Distinguish direct and indirect normativity.02Why does Bostrom favor indirect normativity?03What is coherent extrapolated volition (CEV)?04What open problems does CEV inherit?05Define moral rightness (MR) and Do What I Mean (DWIM), and when each beats CEV.06State the differential technological development principle.07How does differential technological development cash out for AI safety in practice?08What does Bostrom mean by crunch time, and how does he prescribe acting under deep timeline uncertainty?