Economic Policy for AGI
Economic Frameworks for the Age of AGI
Authors:
- Julian Jacobs (Economist and Research Scientist, Google DeepMind)
- Alex Imas (Director of AGI Economics, Google DeepMind)
Introduction: The Looming Economic Shift
There is a palpable and growing sense of apprehension regarding how Artificial Intelligence will reshape the global labor market and broader economy. This anxiety has already permeated political discourse and is particularly acute among younger generations—notably, this fear exists even though before the effects of AI have manifested in most official economic datasets.
The trajectory of AI capabilities has been exponential in recent years, with no indication of a plateau. We are moving toward a future defined by Artificial General Intelligence (AGI): systems capable of executing cognitive functions at a level equal to or greater than the average human.
The Paradox of Progress
Historically, massive technological leaps have followed a specific pattern:
- Long-term: They elevate the general standard of living, improving health and increasing leisure.
- Short-term: They often trigger volatility, including wage stagnation, the collapse of regional economies, and widened wealth gaps.
"The central question we face is: Can we preserve the immense benefits of technological evolution while mitigating or entirely avoiding the human suffering that typically accompanies such transitions?"
Historical Context vs. The AGI Era
In the past, governments were often reactive. During the Industrial Revolution, redistributive policies were typically implemented after the social shock had already occurred. This lag was caused by a combination of primitive data access, weak institutions, and a hesitation to intervene too early.
However, AGI differs from previous innovations in a fundamental way. While earlier tools automated routine tasks, AGI automates intelligence itself.
Comparison of Technological Shocks
| Feature | Industrial Revolution / Early 20th Century | The AGI Transition |
|---|---|---|
| Primary Target | Physical labor & routine cognitive tasks | General cognitive capabilities |
| Adaptation Speed | Slow; policy followed the shock | Potentially faster than human adaptation |
| Labor Impact | Created new industries to absorb labor | Uncertain; could reduce demand for most cognitive work |
| Policy Approach | Reactive/Post-hoc | Proactive/Trigger-based |
Navigating Uncertainty: Three Potential Scenarios
Because the diffusion of AGI is so unprecedented, we cannot be certain of the outcome. We can conceptualize the future through three lenses:
- Scenario A (Minimal Impact): The anticipated economic disruptions and gains simply fail to materialize.
- Scenario B (Traditional Shock): AGI behaves like previous technologies—disrupting specific roles but eventually spawning new occupations, keeping labor a constant force in the economy.
- Scenario C (The Great Break): AGI marks a systemic departure from history, drastically reducing the necessity for human labor across nearly all cognitive fields.
Why there is room for optimism
Despite the risks, we are better equipped today than our ancestors were. We possess:
- Advanced social sciences to model technological shocks.
- More sophisticated tools for measuring labor market shifts.
- A more robust governmental toolkit for scaling interventions.
A Strategy for Flexible Governance
To avoid the binary trap of "acting too late" or "overreacting prematurely," the authors suggest implementing policies tied to empirical_triggers.
def policy_response(economic_data):
if economic_data.unemployment_rate > TRIGGER_THRESHOLD:
deploy_intervention("Universal Basic Income / Retraining")
elif economic_data.wage_growth < STAGNATION_LIMIT:
deploy_intervention("Wage Subsidies")
else:
maintain_monitoring_status("Observation Mode")
The Three Primary Challenges
To realize this flexible approach, society must solve the following:
- Data Granularity: We lack real-time, macro-level data. Platform metrics and small experiments do not equal systemic economic reality.
- Benefit Distribution: We must identify mechanisms that ensure AGI's wealth is shared while maintaining individual human agency.
- Evaluation Rubrics: There is no standardized way to compare different policy interventions across shared societal values.
The Evaluation Framework
To bridge these gaps, a unified framework has been proposed to analyze policies across four key dimensions. The primary goal is to maximize the societal welfare function, which can be simplified as:
Dimension 1: Welfare and Resilience
This focuses on the policy's ability to maintain material living standards and ensure the economy can withstand shocks.
Visualizing the Framework
The following icons represent the core pillars of the proposed evaluation system:
