This project analyzes the Meiji Restoration (1868–1912) to identify the legal, economic, and social mechanisms that gradually displaced Japan's samurai class, then maps those mechanisms to AI-driven displacement risks and develops policy recommendations for preventing gradual human disempowerment.
About the project
Gradual disempowerment describes a distinct failure mode: AI becomes powerful, existing influence-seeking systems (companies, governments, ideologies) adopt AI to gain advantage, and humans become obsolete. Current human alignment with civilization depends on humans having power and utility. This displacement could occur even if society avoids destruction from catastrophic AI risks.
The Meiji Restoration (1868–1912) offers a case study: Japan's samurai class was dismantled through modernizing reforms, many supported by samurai themselves. By the time resistance emerged, the institutional changes were entrenched.
This project maps Meiji displacement mechanisms onto AI-era dynamics to answer the following question: What specific legal, economic, and social mechanisms produced displacement, and which have AI-era analogues? The deliverable is a historically grounded framework for recognizing displacement dynamics and intervening before resistance windows close.
If time allows, further research may cover policies that could have interrupted displacement, and how they may apply to AI-driven disempowerment.
Theory of change
Gradual disempowerment is an underexplored failure mode. Existing work identifies the risk but leaves key questions open: How do we distinguish beneficial AI augmentation from problematic displacement of human influence? What are the tipping points beyond which human influence becomes critically compromised? When do resistance windows close? Therefore, understanding how displacement unfolds is key for timely policy interventions before resistance is no longer possible.
Your role
Mentees will function as lead researchers and primary authors. They will be responsible for conducting the historical research on Meiji-era displacement mechanisms and constructing the mapping to AI-era dynamics. Mentees will contribute the bulk of the work, while I will provide strategic direction on the AI safety framing and overall argument.
Prerequisites
Familiarity with gradual disempowerment and loss-of-control arguments in AI safety Background in history, political economy, law, or policy—ideally at least one upper-level undergraduate or graduate course in economic history, institutional analysis, or comparative political development Research ability to locate, read, and synthesize historical scholarship Analytical writing skills Familiarity with AI safety or AI governance concepts Prior coursework or experience engaging with historical case studies
Bonus (not required): Japanese language proficiency (would enable engagement with primary sources and Japanese-language scholarship) Prior study of Japanese history, particularly the Meiji era or Tokugawa period
Application question(s)
Please read the executive summary of Gradual Disempowerment (https://gradual-disempowerment.ai/).
- In which ways do you think that the scenario the authors describe is likely or unlikely?
- Should mitigating gradual disempowerment be as high of a priority as mitigating other frontier AI risks (e.g., loss of control, adversarial misuse, authoritarian lock-in)? Why or why not?
About the mentor
Alex Mark
ERA:AI
Alex is currently a research fellow with Cambridge ERA:AI. Previously, he was a research fellow with the Cambridge Boston Alignment Initiative. Before pivoting to AI safety, he was a public defender in Los Angeles County and later a housing attorney with Greater Boston Legal Services.