AI Economic Growth: Scenarios, Risks, and Future of Work

Automation has existed for centuries, but the current wave of AI‑driven automation feels distinct from past revolutions. Over the last 150 years, average living standards in the United States have risen at a remarkably steady 2 % per year, even as electricity, the internet, and other general‑purpose technologies reshaped the economy.

Two Scenarios for the Future

Accelerating Growth Scenario

If AI takes over software engineering, it can also generate new algorithms, creating “virtual remote workers” that eventually control physical robots. This cascade of idea generation and automation forms a flywheel effect: each new idea fuels more automation, which in turn produces even more ideas. Simulations of this feedback loop suggest annual growth could reach 30 %, a dramatic departure from historical trends.

Business as Usual Scenario

In the business‑as‑usual path, growth remains anchored at the historical 2 % rate. The reasoning is that ideas become harder to discover over time, so each new general‑purpose technology—AI included—simply replaces the previous one without accelerating the overall pace of improvement.

The “Weak Link” Framework

Production processes often resemble a chain of tasks where the overall output is constrained by the slowest or most difficult step. Even if AI automates 17 of 20 tasks, the remaining three human‑performed tasks act as bottlenecks. This weak‑link dynamic explains why the “good” of abundance takes a long time to arrive: humans continue to perform essential, non‑automated work that limits total output.

Labor Market Implications

AI can raise productivity by handling routine components of jobs, freeing workers to concentrate on high‑value weak‑link tasks. However, if AI eventually automates every task a human can perform, many workers will be pushed into “second‑best” occupations with lower wages. Historical automation has not raised overall unemployment, yet AI’s ability to replace both cognitive and physical labor introduces new uncertainty about the future composition of work.

Catastrophic Risks

A “bad actor” scenario envisions an easily jailbreakable, super‑intelligent model being weaponized by terrorists to design deadly, contagious viruses. An “alien intelligence” scenario warns that creating a super‑intelligence more powerful than humanity could lead to an existential loss of control, as historically dominant entities tend to subjugate less advanced ones. Unlike the gradual emergence of economic abundance, these dangerous outcomes could materialize very quickly.

Finding Meaning in a World of Abundance

In a post‑work economy, leisure, community, and experience become primary sources of meaning. Activities that require a human touch—live sports, music performances, and other arts—retain value precisely because they are performed by people. The distinction between “work as a bad” (paid labor) and “work as a source of meaning” sharpens, suggesting that future fulfillment will hinge on pursuits that cannot be replicated by machines.