💻 computer science

AGI and the Limits of Value Production

This paper presents a political-economy model arguing that under a strict labor theory of value, the near-complete substitution of living labor by AGI will eventually compress the source of surplus value to zero, causing the social rate of profit to collapse and revealing a fundamental limit to value production.

Original authors: Zichen Song

Published 2026-06-05

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Zichen Song

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: A New Kind of Machine

Imagine the economy as a giant factory. For the last 200 years, machines have helped humans work faster. They took over boring, repetitive tasks (like lifting heavy boxes or typing simple letters), but humans were still needed to run the factory, fix the machines, and make the final decisions.

This paper asks a scary question: What happens when the machines get so smart they can do everything?

The author, Zichen Song, builds a mathematical model to see what happens when "Artificial General Intelligence" (AGI) doesn't just help humans, but completely replaces them. The paper argues that under a specific economic theory (Marxist value theory), this total replacement could cause the entire profit system to collapse.

The Two Key Concepts: "Can Do" vs. "Will Do"

The paper makes a very important distinction between two things that people often mix up:

Technical Substitutability (The "Can Do"): This is the ceiling. It's the maximum amount of work the AGI could do if it were perfect. Imagine a robot that is physically capable of cooking every meal in a restaurant.
Actual Adoption (The "Will Do"): This is what the business owner actually chooses to do. Even if the robot can cook, the owner might not hire it yet because it's too expensive, or the staff is hard to fire, or the robot is glitchy.

The Analogy: Think of a video game.

Technical Substitutability is the highest level the game can reach (Level 100).
Actual Adoption is the level you are currently playing at. You might be stuck at Level 10 because you don't have the right gear (money) or you are afraid of the boss (risk).
The paper says: Just because the game can go to Level 100 doesn't mean you will get there tomorrow. But eventually, as the game gets easier (cheaper), you will climb higher.

The Core Rule: Who Makes the Money?

The paper relies on one strict rule: Only living humans create new value (profit).

Humans (Living Labor): They are the "engine" that creates the profit.
Machines (AGI): They are the "fuel tank." They transfer value and make things cheaper, but they don't create new value themselves.

The Analogy: Imagine a bakery.

The baker (human) is the only one who can turn flour into a delicious, valuable cake.
The oven (machine) is necessary to bake the cake, but the oven doesn't "create" the cake's value; it just helps the baker.
If you replace the baker with a robot oven that bakes perfectly, the oven is still just a tool. If there is no baker left to "create" the value, the profit disappears.

The Dangerous Slide: What Happens When We Replace Everyone?

The paper models a transition where AGI gets better and cheaper over time. Here is the step-by-step slide into trouble:

The Cost Drops: As AGI gets smarter, it becomes cheaper to hire a robot than a human.
The Switch: Businesses start firing humans and hiring robots because it saves money.
The Trap: As more humans are fired, the "living labor" (the human engine) shrinks.
- Since humans are the only source of new profit, less human labor means less total profit.
- Meanwhile, the business has spent a lot of money buying all these robots (Constant Capital).
The Result: The business has a massive pile of expensive robots but very few humans to generate profit. The Organic Composition of Capital (the ratio of machines to humans) skyrockets.

The Analogy: Imagine a car race.

The humans are the fuel.
The robots are the car chassis.
If you keep buying bigger, fancier chassis but stop buying fuel, the car stops moving. No matter how good the car is, without fuel, it goes nowhere.

The "Zero" Limit

The paper calculates a "limiting case." This is the point where AGI replaces almost 100% of the work.

If humans are almost entirely replaced, the "fuel" (living labor) runs out.
If there is no fuel, the Surplus Value (profit) drops to zero.
If there is no profit, the Profit Rate drops to zero.

The paper concludes that near-total AGI adoption isn't just a "super-efficient" future; it is a boundary case where the system breaks down because the source of its value (humans) has been removed.

The Timing Gap: Why It Doesn't Happen Instantly

The paper also notes that this doesn't happen overnight. There is a "lag."

Technical Feasibility: The robots become capable of doing the job first.
Actual Adoption: It takes time for companies to actually switch over due to costs and organizational headaches.

The paper shows that during this lag, the economy is in a dangerous spot. The robots are ready, but the humans are still there. As the switch finally happens, the "fuel" (humans) is cut off faster than new jobs can be created to replace them.

The Solution Proposed

The paper suggests that to avoid this collapse, we cannot just let technology run wild. We need to manage the transition:

Create New Jobs: We need to find new areas where humans are needed (creative, social, or complex judgment tasks) faster than robots replace old ones.
Slow Down: Regulate how fast companies can switch to full automation to give the economy time to adjust.
Hybrid Systems: Encourage a mix of humans and machines, rather than total replacement, to keep the "fuel" (human labor) flowing.

Summary

In simple terms, this paper argues: If we replace all human workers with super-smart robots, and if we believe that only humans create value, then the economy will run out of profit.

It's like a factory that fires all its workers to save money, only to realize that without workers, the factory produces nothing of value, and the business goes bankrupt. The paper warns that without careful planning to keep humans involved in the economy, AGI could lead to a system where value production hits zero.

Technical Summary: AGI and the Limits of Value Production

Problem Statement
The paper addresses a theoretical gap in understanding Artificial General Intelligence (AGI) within political economy. Unlike previous waves of automation that replaced specific tasks while leaving large domains of living labor intact or creating new labor fields, AGI possesses the potential to expand the substitution frontier toward near-complete replacement of productive, administrative, analytical, and coordinative activities. The central problem is to model the transition from the initial moment AGI becomes economically capable of replacing labor to the later moment it becomes technically and actually capable of near-complete replacement. The paper seeks to determine whether this transition represents a simple efficiency gain or a structural boundary case for value production under strict political-economy assumptions.

Methodology
The paper develops a dynamic political-economy model built on a strict value-theoretic assumption: living labor is the sole source of new value and surplus value, while AGI-based systems only transfer, reorganize, and cheapen production without creating new value. The model isolates the core dynamic relation between AGI capability, actual adoption, and the remaining quantity of living labor.

Key methodological distinctions include:

Technical Substitutability ( $q(A)$ ): Defined as the maximum share of labor tasks AGI can perform at a given capability level $A$ . This is modeled as an increasing, concave function of capability ( $q'(A) > 0, q''(A) < 0$ ) with a limit of 1.
Actual Adoption ( $\alpha(t)$ ): Defined as the realized share of labor tasks replaced by capital, accounting for cost, profitability, and adoption frictions. This is distinct from technical feasibility; adoption adjusts continuously toward the technical ceiling.
Cost Dynamics: The model assumes unit replacement costs ( $m(A)$ ) decline exponentially as AGI capability rises, creating an economic incentive for adoption once costs fall below the wage rate ( $w$ ).
Capital Structure: The model tracks the organic composition of capital ( $\Omega$ ), defined as the ratio of constant capital (pre-existing + AGI-related) to variable capital (wages).
Dynamic Simulation: The paper employs a continuous-time differential equation to model the evolution of adoption, incorporating a "gate function" that accelerates adoption as replacement costs drop below wages, while acknowledging organizational lags.

Key Contributions
The paper makes three primary contributions to the literature:

Formalization of the Transition: It formalizes the movement from partial to near-complete AGI substitution as a dynamic political-economy process rather than a static comparison, identifying specific threshold times for economic feasibility and near-complete replacement.
Separation of Concepts: It rigorously separates technical substitutability (what is possible) from actual adoption (what is realized). This allows the derivation of separate threshold times for technical feasibility ( $t_{tech}^f$ ) and actual adoption ( $t_{actual}^f$ ), highlighting the existence of an "adoption lag."
Structural Analysis of Value: It demonstrates that under strict value theory, deeper AGI adoption monotonically raises the organic composition of capital and exerts downward pressure on the social profit rate. It identifies a limiting case where near-complete substitution leads to the exhaustion of the living-labor basis of value production.

Results
The model yields several critical analytical results:

Organic Composition: As actual adoption ( $\alpha$ ) increases, the organic composition of capital ( $\Omega$ ) rises monotonically. This occurs because constant capital increases (via AGI investment) while variable capital (wages) decreases.
Profit Rate Pressure: The social profit rate ( $r$ ) is strictly decreasing with respect to actual adoption ( $\partial r / \partial \alpha < 0$ ). This is because the numerator (surplus value) shrinks as the living-labor base contracts, while the denominator (total advanced capital) remains positive due to constant capital.
The Limiting Case: In the limit where actual adoption approaches complete substitution ( $\alpha \to 1$ ) and new labor fields fail to compensate for displaced labor, living labor, surplus value, and the social profit rate all tend toward zero.
Threshold Dynamics: The model identifies two distinct thresholds:
- $t_p$ : The time when AGI replacement cost equals the wage, marking the onset of economic feasibility.
- $t_{tech}^f$ : The time when the technical ceiling reaches a near-complete benchmark (e.g., 99%).
- $t_{actual}^f$ : The time when actual adoption reaches the benchmark.
- The numerical illustration confirms that $t_{actual}^f > t_{tech}^f$ , quantifying the lag caused by adoption frictions.
Living Labor Contraction: Living labor declines when the proportional rate of actual adoption exceeds the proportional growth of new labor fields ( $\dot{\alpha}/(1-\alpha) > \dot{L}/L$ ).

Significance and Claims
The paper claims that near-complete AGI substitution should not be viewed merely as an efficiency transition but as a boundary case for value production. Under the strict assumption that only living labor creates value, the system faces a structural contradiction: as AGI adoption deepens, the source of surplus value (living labor) is eroded while the capital base expands.

The significance of the model lies in its identification of a potential crisis point. If actual AGI adoption persistently outpaces the creation of new labor fields, the system moves toward a state where the profit rate collapses. The paper concludes that the central policy problem is not whether AGI should be adopted, but how its adoption is socially organized. It suggests that sustainable development requires managing the pace of adoption, facilitating the creation of new labor domains, and supporting hybrid human-machine production structures to prevent the exhaustion of the living-labor basis of the economy. The model does not predict an inevitable crisis but highlights the conditions (specifically, the mismatch between adoption speed and labor field expansion) under which such a crisis becomes structurally likely.