Sam Hammond on AI, Techno-Feudalism, and the Future of the State

Episode Summary

Executive Summary: Samuel Hammond argues AI is not just a productivity tool but a regime-changing technology that will reshape state capacity, private institutions, and social organization by 2040. He contrasts three futures—AI Leviathan, techno-feudal fragmentation, or co-evolution—and sees the most likely path as a smaller, more privatized, AI-mediated state with courts, regulation, education, and identity systems increasingly outsourced or augmented by private actors.

Main Topics: AI’s current trajectory and model scaling (Priority: 5/5): The discussion explains how AI evolved from symbolic approaches to deep learning and transformers, and why scaling data, compute, and model size has produced qualitative leaps like GPT-4 and multimodal systems. AI safety and policymaking on Capitol Hill (Priority: 4/5): Beckworth and Hammond discuss the surge of Senate hearings, congressional attention, and the mismatch between public awareness, policy urgency, and the speed of technical change. AI, consciousness, and sentience (Priority: 3/5): Hammond lays out a computational functionalist view of mind, argues AI could potentially develop or appear to develop subjective experience, and explores the philosophical implications of increasingly agentic models. Historical analogies: printing press, internet, and regime change (Priority: 5/5): He compares AI to prior general-purpose information technologies that triggered institutional shifts, using the English Civil War, the Arab Spring, and China’s surveillance response as models for indirect second-order effects. Institutional economics and the shrinking transaction costs of AI (Priority: 5/5): AI lowers monitoring, bargaining, and search costs, which could weaken traditional roles of firms and governments and push more governance into private, AI-enabled institutions. Three future state models: Leviathan, fragmentation, or co-evolution (Priority: 5/5): Hammond argues the likely path is techno-feudal fragmentation, with a smaller state focused on core competencies, while acknowledging a possible middle path of constrained Leviathan and institutional adaptation. Timeline to 2040 and labor/content disruption (Priority: 4/5): He outlines a timeline in which AI reaches human-level task replication around 2029, synthetic content dominates online media, and robots scale into manufacturing by the late 2030s, transforming jobs and culture.

Key Arguments: AI’s most important effects will be indirect and institutional, not just direct misuse like deepfakes or malware. The printing press and internet show that information technologies can trigger regime change by altering coordination, censorship, and political mobilization. AI reduces transaction costs across monitoring, bargaining, and information search, undermining the need for some functions currently performed by firms and government. Private institutions will likely adapt faster than public agencies because they face fewer veto points, faster procurement cycles, and stronger incentives to innovate. A Chinese-style AI Leviathan is possible, but in liberal societies the more likely outcome is fragmentation and the expansion of private governance. Courts and agencies are already overloaded; AI lawyers, AI auditors, and AI case review could push dispute resolution and compliance toward private or semi-private systems. By the end of the decade, AI may be able to emulate humans on many mental tasks, making a broad labor shift feel closer than most people assume. Mass culture may fragment further as AI enables highly personalized media, entertainment, and information ecosystems. Human value may shift toward identity, authenticity, and personal brand as synthetic content becomes abundant. A constrained, high-capacity state is preferable to both despotism and anarchy, but current U.S. administrative and procurement structures may struggle to keep up.

Data Points: Senate hearings on AI: 4-5 hearings in one week - Hammond says Capitol Hill held multiple AI and adjacent-topic hearings in the prior week. Schumer AI listening sessions: 9 sessions planned this calendar year - He describes a Senate effort to hold recurring six-hour AI briefings. ChatGPT/GPT leap: GPT-4 passed the bar exam - Used to illustrate qualitative gains from scaling compared with GPT-2 and GPT-3. Gemini size relative to GPT-4: 5-10 times bigger - Rumored scale of Google’s upcoming multimodal model. Next-generation model scale: 1,000x bigger than GPT-4 - Hammond cites companies training models expected next year. NYC taxi market shift: 90% taxis to 10% taxis - Ride-hailing example of a rapid institutional regime change in urban transport. AI-generated code on GitHub: Over half of new code - Used to show AI already acts as a coding copilot. FTC healthcare division staffing: 36-person team - Example of a small government team overseeing a large industry. Email discovery burden: 40,000 emails - Illustrates how AI could accelerate regulatory review and enforcement. NVIDIA GPU production: More GPUs this year than cumulative history - Used as a proxy for compute buildout supporting AI scaling. Cloud computing growth: Double or triple in 3-4 years - Projected expansion of global cloud infrastructure. Cloud compute CAGR: 30-40% - Hammond cites growth in available compute as a key driver. Human-level task emulation target: Around 2029 - Based on scaling laws and forecasts from Epoch AI. AI content timeline: By the 2030s, much online content synthetic - He argues entertainment and media will be increasingly AI-generated. Robotics/manufacturing milestone: 2036-2039 - Projected period for general-purpose robots in large-scale manufacturing.

Pivotal Quotes: "AI safety is usually conceptualized in terms of what AI will do directly, rather than in terms of AI's likely indirect second order effects on society, and the shape of our institutions. This is an enormous blind spot." — Samuel Hammond: He argues policymakers focus too narrowly on direct harms and miss structural institutional change. "There are really three possible futures." — Samuel Hammond: He introduces his framework: AI Leviathan, techno-feudal fragmentation, or co-evolution/constrained Leviathan. "The state as we know it could be very different by 2040." — David Beckworth: Beckworth summarizes the scale and relevance of Hammond’s thesis about institutional transformation.

Implications: Listeners should expect AI to reshape regulation, courts, media, labor, and identity faster than governments can adapt. The biggest risk is institutional mismatch: either heavy surveillance or fragmented private governance unless public institutions modernize quickly.