Feature
As transformative as electricity — and perhaps how we consume electricity — artificial intelligence (AI) may shock the economy in unpredictable ways. Economists are split, and Nobel laureates diverge — is a four-day work week or a deepening chasm of inequality imminent? Maybe both.
Here, we’ll look at five timely perspectives on AI from a variety of economists. Unlike tech executives, who often envision expansive potentiality, the economists featured below have a more measured stance on the rate of impact from AI — depending, of course, on efficacy, adoption and any emergent capabilities of future frontier models.
Paul Krugman: Productivity Gains From Technology Require New Approaches
A New York Times opinion columnist and Nobel Prize winner, Paul Krugman notes that even with significant technological advances, know-how and infrastructure changes mean productivity often lags technology. Job losses are inevitable along with these changes and can’t be stopped. AI may change everything but not imminently.
Krugman has a reputation for underestimating the impacts of technology and, in turn, emphasizes he is an economist not a technologist. From an economist perspective he has highlighted that there is an inherent challenge of measuring the impacts of technology development on productivity. “Technology is often the catch-all for what can’t be measured in how economists (and governments) measure economic growth,” Krugman said.
Speaking of governments and their debt, Krugman sees AI gains likely to offset any concerns around growing governmental debt.
Yanis Varoufakis: AI Robot Taxes Can Reduce Rent-Seeking Tech 'Fiefdoms'
The former economic minister of Greece during its debt default crisis, Yanis Varoufakis is intimately familiar with government debt. He has since become known for his analysis and criticism of technology companies. Varoufakis thinks that “capitalism is dead," having been replaced by something he terms “technofeudalism.” Within this economic paradigm, private economic “fiefdoms” are overseen by tech giants, like Google, Meta and Amazon, where they have highly controlled rent-seeking marketplaces. He sees AI as turbocharging and accelerating the critical battleground for behavior and economic control of internet users.
Varoufakis regularly returns to the need for a universal basic income, in which all citizens receive a monthly stipend. A robot tax levied on tech giants, such as Amazon’s rapidly growing robot labor force, may be an effective source to fund these programs. Even within these concerns, however, Varoufakis is not a technophobe — he sees AI as a brilliant human achievement.
On a more practical level, Varoufakis also notes the ability of AI to foundationally affect traditionally understood economic perspectives, such as labor unions. He notes how Amazon’s AI-monitored warehouse cameras, ostensibly for operational efficiency, will also serve as an effective means to detect, anticipate and curtail labor-related activities.
Tyler Cowen: AI is the New Economic Growth Engine
Professor Cowen of Geroge Washington University is a regular Bloomberg columnist and argued through his “Great Stagnation” theory that American economic growth had peaked. However, with the proliferation of AI, Cowan sees a shift: “We are now in the early stages of a revolution that will change everything about our world,” Cowan said.
While the Great Stagnation may be “over,” Cowan notes that some AI innovations — such as longevity or medicinal innovations — may not affect GDP. Cowan predicts a one-quarter to one-half of a percentage point GDP impact from AI annually. Similar to Krugman, Cowan highlights the importance of other constraints, like know-how, infrastructure and technology.
While today’s coverage of AI often wonders whether AI is peaking, Cowan sees the consistent and accelerating product launches as evidence of a broad shift and steady growth. At his regular blog, he often analyzes media and other meta-commentaries, and he highlights the importance of contextualizing media narratives and definitions, such as the term artificial general intelligence (AGI):
“Five years ago, if people had seen Claude 3 Opus, would they have thought we had AGI? Just as a descriptive matter, I think the answer to that question is yes,” Cowan says.
Given the importance of AGI to economic and socio-political prognostications, like many of the economic studies within this article, Cowan’s comment is especially prescient. AGI lacks a clear definition, and there are practical reasons why there may not be consensus that AGI has been reached — a notable one being that OpenAI’s AGI IP may be outside of OpenAI’s scope of relationship with Microsoft.
Anton Korniek: Speed of Change Determines Future Wages
The arrival of AGI, however defined, generally means an end to work as currently understood. Professor Korniek of the University of Virginia modeled what the economic impacts may look like if the widespread changes of AGI occurred, as predicted by AI boosters. In a working paper co-written with Donghyun Suh, he examines situations in which AGI is achieved at different time scales, five years and 20 years. In the paper and subsequent interviews, he highlights several key considerations for AI economy growth:
- Task complexity: Is there an upward limit of tasks that only humans can perform?
- Speed of adoption: How quickly does AI automation occur?
- Comparative advantage: How much will trade disparities and balances be affected?
If there is no upward limit of tasks that are human only, then wages and capital can both grow. However, if all possible work can be done by AI, then wages will collapse, and capital will dramatically accumulate. Trade may also become less sustainable if comparative advantages become rapidly commoditized. Consistently, the speed of change is a key aspect of future predictions in their model.
While the rate of AI change is uncertain, Korniek highlights the need for policy action to address economic concerns:
“Policymakers are not quite there yet — that’s my concern. I'm nervous about this, because we may not have all that much time,” Korniek says.
In a written statement to the U.S. Senate AI Insight committee, Korniek highlights the importance of the government actively developing a commission to investigate likely labor-related scenarios as well as developing plans for potential overhauls of social safety and tax code if AGI arrives quickly.
Daron Acemoglu: AI Presents a Clear Choice of Economic Outcomes
The rapid approach of AI requires decisive decision making. Professor DaronAcemoglu of MIT, one of the most prolific and cited modern economists, sees AI as presenting a profound choice of outcomes between shared prosperity or authoritarian inequality. Broad reform of business models, government safety nets and higher involvement of labor are essential. In a prepared statement to the U.S. Senate, Acemoglu highlighted that regulations and tax code without considerations of software and AI automation within the U.S. are pushing the labor economy in a “worrying direction.”
Acemoglu suggests that AI may simply replace jobs without creating new ones, echoing Korniek. Acemoglu also highlights the opportunity and need for AI to focus on developing useful solutions versus “so-so” automation. For example, AI tools could be used to solve cancer or revolutionize education and training — or more effectively reject medical insurance claims. Without widespread AI adoption and equitable access, rent-seeking behaviors could displace gains from innovation. The crux of Acemoglu’s argument is that the paradigm-shifting abilities of AI also require that business and regulation change in a similarly significant fashion.
Beyond Impacting the Economy, AI is Likely to Impact Economics
While economists see impacts from AI across a variety of different arenas, it is likely to impact the study of economics as well. Beyond potential changes to orthodoxies on labor, unions and productivity, the machine learning (ML) techniques of AI are already impacting the study of economics itself. Professor Esther Duflo of MIT extensively used machine learning for her Nobel Prize-winning research on poverty, and Korinek similarly emphasizes the opportunities of using generative AI for economic analysis.
This meta AI instance — where AI simultaneously affects the economy and economist’s ability to measure the economy effectively — suggests that each new model launch may correspond with a change in economic prognosis and definitions. Perhaps a model launch accelerates calls for UBI, spurs rapid growth or simply continues the as-is trajectory.
When’s the next major discourse shift in economics? Perhaps the expected launch of OpenAI latest GPT model this year.
