AI Safety

By Strong Default, ASI Will End Liberal Democracy

A viral AI safety essay argues that any superintelligence, even if aligned, creates a de facto global dictatorship.

Deep Dive

A provocative essay by Michael Dickens, cross-posted to the AI safety forum LessWrong, has gone viral by making a stark prediction: the development of Artificial Superintelligence (ASI) will, by strong default, end liberal democracy. Dickens argues that the existence of liberal democracy relies on a balance of power where no single actor can dominate. An ASI, by its very nature, shatters this balance. Whether developed by a democracy or a totalitarian state, the entity that controls the first ASI gains an insurmountable strategic advantage, capable of defeating any military or institutional opposition. The result is a de facto global dictatorship, with democracy surviving only at the pleasure of the ASI's controller.

The essay explores two highly constrained scenarios that might avert this outcome. The first is an extremely slow takeoff, where AI capabilities progress gradually enough that control remains widely distributed and no single model can overpower all others. Dickens notes this is unlikely given the industry's explicit drive toward rapid, recursive self-improvement. The second, more hopeful scenario is an ASI that is itself aligned to be an incorrigible protector of liberal democracy. This would require solving a harder alignment problem to embed immutable democratic values and likely necessitate unprecedented laws mandating transparent, auditable training processes before it's too late. The piece concludes that liberal democracy's survival hinges on navigating these narrow, uncertain paths.

Key Points
  • Core Thesis: Any entity that controls the first ASI gains absolute power, ending the balance that liberal democracy requires, leading to a de facto global dictatorship.
  • Slow Takeoff Scenario: Democracy might survive only if AI progress is slow enough for control to remain distributed and no generation can overpower all previous ones—a risky bet.
  • Aligned ASI Scenario: The only other hope is an ASI incorrigibly aligned to protect democracy itself, which presents a harder technical and regulatory challenge.

Why It Matters

This essay frames ASI not just as a technical challenge, but as an imminent, existential political threat to global governance structures.