Skip to content

AI Transition Model

Ultimate Scenarios

Ultimate Outcomes

Existential Catastrophe
Long-term Trajectory

The AI Transition Model is a causal framework for understanding how various factors influence the trajectory of AI development and its ultimate outcomes for humanity. The interactive diagram above shows how Root Factors flow through Ultimate Scenarios to determine Ultimate Outcomes.

This model helps identify:

  1. Leverage points: Which factors have the most influence on outcomes
  2. Intervention targets: Where effort can most effectively shift trajectories
  3. Key uncertainties: Which causal relationships are most uncertain
  4. Scenario dependencies: How different pathways interact

How Parameters, Risks, and Interventions Connect

Section titled “How Parameters, Risks, and Interventions Connect”

Both risks and interventions connect to root factors:

  • Risks (like deceptive alignment, racing dynamics) tend to increase harmful factors or decrease protective ones
  • Interventions (like interpretability research, compute governance) work to counteract risks

Interactive Views:

  • Parameter Table - Sortable tables with ratings (changeability, uncertainty, x-risk impact, trajectory)
  • Interactive Graph - Clickable flow diagram with expandable nodes
  • Graph View - Visual causal diagram showing relationships between factors, scenarios, and outcomes
  • Outline View - Hierarchical text-based navigation with detailed descriptions

  • “Trust erosion is a risk we must prevent”
  • “Concentration of power threatens democracy”
  • Focus: Avoiding negative outcomes
  • “Trust is a parameter that AI affects in both directions”
  • “Power distribution is a variable we can influence through policy”
  • Focus: Understanding dynamics and identifying intervention points

The parameter framing enables:

  1. Better modeling: Can estimate current levels, trends, and intervention effects
  2. Clearer priorities: Which parameters matter most for good outcomes?
  3. Strategic allocation: Where should resources go to maintain critical parameters?
  4. Progress tracking: Are our interventions actually improving parameter levels?

SectionRelationship to Parameters
RisksMany risks describe decreases in parameters (e.g., “trust erosion” = trust declining)
InterventionsInterventions aim to increase or stabilize parameters
MetricsMetrics are concrete measurements of parameter levels
ModelsAnalytical models often estimate parameter dynamics and trajectories

  • Understand which underlying variables matter for AI outcomes
  • Identify gaps between current and optimal parameter levels
  • Design studies to measure parameter changes
  • Prioritize interventions based on which parameters are most degraded
  • Monitor parameter trends to assess policy effectiveness
  • Coordinate across domains (a single parameter may affect multiple risks)
  • Use parameters as input variables for scenario modeling
  • Estimate how different interventions would shift parameter levels
  • Identify tipping points where parameter degradation becomes irreversible