What impact will AI have on jobs and the economy? (with Anton Korinek)
Professor Anton Korinek discusses how AI could be as transformative as the Industrial Revolution, potentially shifting economies from labor to capital, and its implications for economic growth, labor markets, and wealth distribution. He explores the potential for a 'growth explosion' and societal challenges.
Deep Dive Analysis
19 Topic Outline
AI's Impact Compared to Industrial Revolution
Labor vs. Capital in Economic Production
Defining Economic Output and GDP
Critiques of GDP as a Welfare Measure
AI's Potential Impact on GDP Per Capita
Arguments for Low vs. High AI Economic Impact
Understanding Exponential Growth and Compounding
Feedback Loops Driving AI Growth Explosion
Aggregate Demand Shortfall and Recession Risk
Critiques Against High AI Growth Projections
Diminishing Returns to Intelligence Argument
AI's Impact on Income Distribution and Wages
Automation's Effect on Specific Professions
AI and Non-GDP Welfare Effects
Limitations of Current Economic Models for AI
Cost of AI Development and Scaling Challenges
Pace of AI Adoption and Organizational Restructuring
Future Taxation in an AI-Transformed Economy
Government Preparedness for Transformative AI
7 Key Concepts
Factors of Production
In economics, these are the main inputs used to produce goods and services, traditionally labor (human workers) and capital (machines, factories). AI is expected to blur this distinction by substituting for labor within the capital category.
Production Functions
Economic models that describe how a given amount of capital and labor can be combined to produce a certain amount of output (GDP). These functions are expected to change fundamentally with advanced AI systems that can substitute for labor.
Productivity Gain
A measure of how much more output can be obtained for a given amount of input. For example, if AI can perform a task for 95% less cost, it represents a 20x productivity gain, meaning the same output can be achieved with significantly less input.
Recursive Self-Improvement
A state where AI systems can improve themselves, leading to a feedback loop where better AI creates even better AI. This process could potentially trigger an exponential 'growth explosion' in the economy.
Aggregate Demand Shortfall
A potential economic issue where widespread job automation and reduced labor income lead to less consumer spending. This could cause a recession if productivity gains don't diffuse widely enough to offset the loss in purchasing power.
Decreasing Returns to Scale
The idea that beyond a certain point, increasing an input (like intelligence) yields progressively smaller increases in output. This is a counter-argument to the idea of an unbounded AI-driven growth explosion, suggesting limits to how much more intelligence can improve productivity.
Complementary vs. Substitute Labor
When AI acts as a substitute for human labor, it can displace workers. When AI acts as a complement, it enhances human productivity, making human labor more valuable and potentially increasing wages.
11 Questions Answered
AI is expected to be as significant, if not more so, by fundamentally shifting the economy from being primarily labor-based to capital-based, similar to how the Industrial Revolution shifted from land to labor.
Labor refers to human workers, while capital encompasses everything else used in production, such as machines, factories, and infrastructure. Currently, labor receives about two-thirds of economic output, but AI could significantly shrink this share.
GDP measures the total value of goods and services produced and is a good indicator of material welfare, but it doesn't capture non-material aspects like love, environmental quality, or the distribution of wealth, and can even include 'defensive expenditures' that don't improve well-being.
While many expect AI to increase GDP per capita, there's debate; some economists predict minimal impact (e.g., 0.07% annual increase), while others foresee a 'growth explosion' due to recursive self-improvement.
Small annual productivity gains, even 1-2%, lead to significant economic growth over time due to compounding, meaning the economy can double in size over a few decades.
AI is almost certain to reduce the labor share of the economy. While overall wages could still rise if productivity gains are large enough, specific sectors or individual workers whose jobs are substituted by AI might experience wage declines or job loss.
Yes, if AI automates only a portion of a job (e.g., 10%) and significantly improves the quality or efficiency of the service, it can make human work more valuable and increase demand for that profession.
The strongest argument is that there may be strongly decreasing returns to scale for intelligence, meaning that beyond a certain point, even vastly superior AI might not yield proportionally higher productivity gains, limiting its economic impact.
Yes, in the short term, a wave of automation leading to widespread job losses and reduced income for white-collar workers could create an aggregate demand shortfall, potentially causing a recession before the full productivity gains are realized and diffused.
AI can have both significant positive impacts (e.g., instant medical advice, improved quality of life) and negative impacts (e.g., misinformation, automated manipulation, increased authoritarian control), many of which are not directly reflected in GDP.
If the labor share of the economy significantly declines, governments may need to shift taxation away from labor income towards consumption or, eventually, capital to maintain funding and address distributional challenges.
5 Actionable Insights
1. Focus on Complementary Skills
Develop skills that complement AI rather than compete with it. If humans remain essential for tasks like final approval or providing the ‘human touch’ in AI-driven processes, their value and wages can significantly increase.
2. Hedge Against AI Uncertainty
Acknowledge the significant uncertainties in AI’s economic impact, including the timing of productivity gains versus job losses. Be prepared for both optimistic and pessimistic scenarios rather than being blinded by optimism.
3. Anticipate Job Disruption
Recognize that AI automation can lead to job displacement, particularly in roles like call service centers or even specialized white-collar professions. Be prepared for potential career transitions or retraining to adapt.
4. Prepare for Economic Shifts
Understand that AI may fundamentally shift the economy from being primarily labor-based to capital-based, potentially shrinking the labor share. This requires rethinking personal financial strategies and societal structures.
5. Rethink Future Tax Systems
Consider that if the labor share of the economy declines significantly due to AI, governments may need to shift taxation away from labor income towards consumption or capital to maintain funding and address distributional challenges.
5 Key Quotes
If you look at how output is produced in the economy, you have what we economists call the two main factors of production, labor and capital. The labor, that's us workers. The capital is everything else.
Anton Korinek
Yeah, you could imagine two societies that are exactly equivalent. They produce the same amount of GDP. But one of them, everyone's miserable because, let's say, like it's just a really psychologically unhealthy culture where everyone thinks that they're worthless unless they work a lot and they never can be good enough. And in another culture, everyone works motivated by love and connection and happiness. And they both produce the same GDP. But like one's a shitty culture and one's a good culture, right? So clearly it's not all that we care about.
Spencer Greenberg
The thing about exponential growth is small numbers compound over time.
Anton Korinek
If there are substantive gaps in things only humans can do, then humans will be very complementary and wages will rise significantly.
Anton Korinek
In economics, history is a better guide to the future than science fiction.
Anton Korinek