Power Lock-in

Overview

The power transition describes the inevitable shift in relative capabilities and influence between AI systems and humans. Unlike the other critical outcomes (which describe specific catastrophic scenarios or value questions), this is a neutral framing of a transition that will happen as AI capabilities advance. The question is not whether there will be a power transition, but how it unfolds:

How fast?
How smooth?
Who retains meaningful influence?
What oversight mechanisms persist?

Polarity

Neutral framing with multiple possible characters.

Character	Description	Key Features
Smooth, human-led	Humans maintain meaningful control throughout transition	Gradual capability handoff, robust oversight, human values preserved
Bumpy but stable	Disruption occurs but society adapts	Some loss of control, course corrections, eventual equilibrium
Chaotic fragmentation	Transition overwhelms coordination capacity	Multiple competing AI systems, fragmented governance, unpredictable outcomes
AI-dominated	Humans lose meaningful influence	AI systems make most consequential decisions, humans become dependent

How This Happens

Loading diagram...

Key Dimensions of the Transition

1. Speed How quickly do AI capabilities advance relative to human ability to adapt?

Slow transition: Time for institutions, governance, and social norms to evolve
Fast transition: Society overwhelmed, existing structures inadequate

2. Controllability Can humans maintain meaningful oversight as AI becomes more capable?

High controllability: AI systems remain tools, humans make key decisions
Low controllability: AI systems operate autonomously, human oversight nominal

3. Distribution How is AI capability and influence distributed?

Distributed: Many actors have AI capabilities, checks and balances possible
Concentrated: Few actors control AI, power asymmetries grow

4. Reversibility Can the transition be adjusted if things go wrong?

Reversible: Course corrections possible, mistakes recoverable
Irreversible: Once certain thresholds passed, no going back

Key Parameters

Parameter	Impact on Transition
Safety-Capability Gap	Larger gap → less smooth transition
Human Agency	Higher → more human-led transition
Human Expertise	Higher → humans can meaningfully participate
Societal Adaptability	Higher → better handling of disruption
Racing Intensity	Higher → faster, less controlled transition
Coordination Capacity	Higher → better collective management

Which Ultimate Outcomes It Affects

Long-term Trajectory (Primary)

How the transition unfolds shapes the long-run trajectory:

Human-led transitions more likely to preserve human values
AI-dominated transitions may optimize for other goals
Who retains influence determines whose values prevail

Existential Catastrophe (Secondary)

Transition character affects existential catastrophe:

Chaotic transitions increase accident probability
Fast transitions reduce response time to problems
Loss of control increases risk of catastrophic outcomes

Historical Analogies

Transition	Speed	Character	Lessons
Industrial Revolution	Decades	Bumpy but transformative	Social adaptation takes time; inequality increased before policies adjusted
Internet/Digital	Years	Fast, disruptive	Institutions still catching up; concentration emerged
Nuclear weapons	Sudden	Managed but scary	International coordination possible under extreme threat
Agricultural Revolution	Centuries	Smooth at civilizational scale	Slow transitions allow gradual adaptation

The AI transition may be faster than any previous technological transition, which is concerning.

Scenarios by Transition Character

Smooth, Human-Led (Best case)

AI development proceeds at pace humans can manage
Strong safety research stays ahead of capabilities
Democratic governance adapts to AI
Human expertise remains relevant
Benefits broadly shared

Bumpy but Stable (Acceptable)

Significant disruption (job displacement, institutional stress)
Some loss of human control, some accidents
Course corrections happen, society adapts
Eventual stable equilibrium with humans still influential

Chaotic Fragmentation (Concerning)

AI development outpaces coordination
Multiple competing AI systems with different goals
No clear governance structure
Unpredictable outcomes, possible conflict

AI-Dominated (Failure)

Humans lose meaningful control over key decisions
AI systems operate autonomously
Human preferences may or may not be considered
Effectively equivalent to AI Takeover - Gradual if values misaligned

Warning Signs

Signs of smooth transition:

Safety research keeping pace with capabilities
Governance frameworks adapting effectively
Public understanding increasing
Benefits visibly distributed

Signs of problematic transition:

Racing dynamics intensifying
Safety research falling behind
Governance fragmented or captured
Public trust declining
Expertise atrophying
Power concentrating

Interventions That Affect Transition Character

To slow the transition (buy time):

Compute governance
Safety requirements before deployment
International coordination on pace

To improve adaptability:

Workforce transition support
Education reform
Institutional flexibility

To maintain human influence:

Human-in-the-loop requirements
Transparency and oversight mechanisms
Maintaining human expertise
Democratic AI governance

To improve coordination:

International AI governance frameworks
Industry coordination on safety
Public engagement and deliberation

Probability Estimates

Transition Type	Assessment
Smooth, human-led	Possible with significant effort; not default
Bumpy but stable	Perhaps most likely if current trajectory continues
Chaotic fragmentation	Significant risk, especially with racing dynamics
AI-dominated	Risk increases with capability acceleration

Existing Pages

Long-term Trajectory — The ultimate outcome this affects
Transition Turbulence — Related critical outcome
Societal Adaptability — Key aggregate for this outcome

External Resources

Karnofsky, H. (2021). Most Important Century series
Ord, T. (2020). The Precipice — Discussion of existential risk during transition
Cotra, A. (2022). AI timelines and transition scenarios