Skip to content

Ultimate Outcomes

Ultimate Outcomes represent what we fundamentally care about when thinking about AI’s impact on humanity. Unlike Scenarios (which describe intermediate scenarios) or parameters (which measure specific factors), Ultimate Outcomes describe the final states we’re trying to achieve or avoid.

There are two Ultimate Outcomes:

  1. Existential Catastrophe — Does catastrophe occur?
  2. Long-term Trajectory — What’s the expected value of the future?

Loading diagram...
OutcomeQuestionKey Ultimate Scenarios
Existential CatastropheDoes civilization-ending harm occur?AI Takeover, Human-Caused Catastrophe
Long-term TrajectoryWhat's the quality of the post-transition future?AI Takeover, Long-term Lock-in

Previous versions of this framework had three outcomes (including “Transition Smoothness”). We moved to two because:

  1. Transition Turbulence is a pathway, not endpoint: How rough the transition is affects both existential catastrophe and long-term trajectory. It belongs in Root Factors.

  2. Cleaner analytical structure: Two outcomes are genuinely orthogonal:

    • You can have low existential catastrophe but poor long-term trajectory (safe dystopia)
    • You can have high existential catastrophe but good conditional value (high-stakes gamble)
  3. Temporal clarity: Existential Catastrophe is primarily about the transition period; Long-term Trajectory is about what comes after.


These outcomes are partially independent—you can have different combinations:

ScenarioExistential CatastropheLong-term TrajectoryExample
Best caseLowHighAligned AI, smooth transition, flourishing
Safe dystopiaLowLowNo catastrophe but authoritarian lock-in
High-stakes successHigh (survived)HighNear-misses but good outcome
ExtinctionVery HighN/ACatastrophe occurs

This independence means:

  • Different Ultimate Scenarios affect different Ultimate Outcomes
  • Trade-offs exist: Some approaches that reduce existential catastrophe might worsen long-term trajectory (e.g., authoritarian control)
  • Both matter: We shouldn’t sacrifice one entirely for the other

How Ultimate Scenarios Flow to Ultimate Outcomes

Section titled “How Ultimate Scenarios Flow to Ultimate Outcomes”
Loading diagram...

Each ultimate scenario has sub-variants with different probability estimates. See the Ultimate Scenarios section for details.


These outcomes map to different phases of the AI transition:

PhasePrimary ConcernRelevant Outcome
Pre-transformative AI (now)Building capacity, avoiding racingExistential Catastrophe (preparation)
Existential Catastrophe PeriodSurviving the transitionExistential Catastrophe
ResolutionHow it resolvesBoth
Long-run TrajectoryQuality of the futureLong-term Trajectory