I’ve written a lot in these newsletters about how artificial intelligence systems can entrench bias. But there’s more to worry about with AI than just bias. AI can replace jobs, undermine human dignity, and transform our economies.
Economist and Nobel laureate Joseph E. Stiglitz has been talking recently about the ways in which AI can be harmful to societies and economic innovation. So, I caught up with him via email to ask him to elaborate on the economic harms that he is worried about with AI.
That AI is replacing labor is just one of many harms. Stiglitz is also concerned that companies are using AI to maximize profits by exploiting customers, and he’s worried that a handful of companies are hoarding what should be public knowledge. He proposes an AI future that is optimized for societal well-being rather than profits.
Stiglitz is a professor at Columbia University and a recipient of the Nobel Memorial Prize in Economic Sciences (2001) and the John Bates Clark Medal (1979). He is the co-chair of the High-Level Expert Group on the Measurement of Economic Performance and Social Progress at the OECD and the chief economist of the Roosevelt Institute. He is a former senior vice president and chief economist of the World Bank and a former chairman of the U.S. Council of Economic Advisers.
Known for his pioneering work on asymmetric information, Stiglitz focuses his research on income distribution, climate change, corporate governance, public policy, macroeconomics, and globalization. He is the author of numerous books, including “People, Power, and Profits: Progressive Capitalism for an Age of Discontent” (2019).
His written responses to my questions, edited for clarity, are below.
Angwin: You have spoken about the need to ensure that AI creates value and increases social welfare. Can you tell me more about your position?
Stiglitz: As an economist, of course, a first cause of both concern and potential celebration is AI’s productivity. It enables us to do more things with fewer resources. In a fundamental sense, and in the aggregate, it could make us richer.
The very fact that AI can enlarge the national pie means that everyone could get a bigger slice. But often that’s not how our economic and political system works. It’s certainly not how it has been working for the past four decades in the era of neoliberalism. Virtually all the gains of increased productivity have gone to the top, while the middle has stagnated and many at the bottom have seen both incomes and opportunity wane.
AI is a transformative technology, and anything this transformative is likely to also reshape not just the economy, but also society. The question of how it transforms society is not inevitable—it depends on how we shape and respond to AI.
Angwin: Related to productivity, there has been much debate about how AI will impact labor. What is your take?
Stiglitz: Let’s look at the Industrial Revolution. That was transformative. It unleashed advances that have enormously increased our standards of living and life expectancies, but we should remember that the initial decades were not pretty. The lives of many ordinary people became worse, as so vividly portrayed in Dickens, for instance. This led to movements like the Luddites, workers who resisted innovation.
There’s a widespread belief that today our inability to manage the smaller-scale changes in technology and globalization has contributed to the polarization of our societies, threatening democracy in two countries that were at the forefront of the Industrial Revolution, the U.S. and France. The changes that AI may bring about are potentially more profound.
In certain areas it has the ability to replace workers. It may be so efficient as to even reduce capital requirements. Resources thereby freed could be re-devoted to other uses. Now, and for the foreseeable future, there are massive needs of society. Infrastructure, health, education, caring for the elderly, research, and, most critically, resources needed for the green transition. Some say we can’t afford the green transition; it will require massive investment and much labor. AI could help free up the resources we need.
That’s the upside, but there is a downside. A decreased demand for labor could result in lower wages, more inequality, and more societal polarization, with all the political problems attendant on that.
Angwin: Can you speak to how reliance on AI may impact competition?
Stiglitz: First, with data functioning as the critical input and a scarce resource, companies will likely “hoard” their data. This becomes a vicious circle that creates a less competitive economy. Data can be used to extract consumer surplus by charging different customers different prices. Economists refer to this as rent extraction. Companies that prosper are not those that are most efficient and that do the best job satisfying customers but those that are best at exploitation, at extracting this consumer surplus. In doing so, inequality will increase and real innovation will diminish.
Second, the basis of the efficiency of the market economy is that everyone faces the same prices—you may remember from your elementary economics courses that marginal benefits should equal marginal costs for all consumers and producers, and that is ensured when everyone pays the same price. Now, AI is used to price discriminate, meaning different people pay different prices, which undermines the foundations of the efficiency of the market economy.
Third, key advances in knowledge have always been placed in the public domain. Knowledge is what economists call a public good. Thomas Jefferson put it well when he said that knowledge is like a candle: When one candle lights another, it doesn’t diminish the first. That’s why basic research has to be funded publicly, and knowledge should be disseminated as widely as possible. While that was true for the basic research underlying AI, an increasing fraction of critical AI research is being done within private organizations, without the full knowledge-sharing that was so important, for instance, in the development of the mRNA vaccines.
Angwin: Do we need regulation to prevent some of the risks of AI?
Stiglitz: Economic outcomes are driven not just by the laws of economics and the laws of nature but also by the laws of man, and these are shaped by the political process. The current economic framework has extended and entrenched market power, undermining the efficiency of the economy. This is a central theme in my recent book “People, Power, and Profits.”
We need regulation—every game needs rules and referees. AI is a new game, and it needs new rules. We should begin with transparency and accountability. For social media, that entails accountability for disinformation (the repeal of Section 230). Europe’s Digital Services Act (DSA) is a good start, as it would rein in a range of digital harms, proscribing some of the nefarious practices. Additionally, Europe’s Digital Markets Act (DMA) will hopefully lead to a more competitive marketplace. AI and the AI marketplace are evolving rapidly, and the regulations that are required will have to evolve rapidly as well.
Angwin: Is there an example you can think of where AI could be re-purposed or improved to better promote social welfare?
Stiglitz: Yes, consider the harms from social media. By now, these are well documented—how hate crimes have been instigated in Myanmar and elsewhere, anxieties among teenagers in the U.S., the political polarization observed in many countries, the shortening of attention spans, vaccine hesitancy exacerbated by disinformation spread through social media. The heightened concern about mis- and disinformation is understandable and well-deserved.
Much of this is not inevitable. The social media companies have shown that they can moderate content when they want to—perhaps not perfectly, but significantly. To be sure, it’s costly, but these companies, which include those among the wealthiest in the world, can hardly claim they don’t have the resources. They lack the desire.
And that’s the critical point: Their objective is to maximize engagement, and many of the adverse effects just described are side products of AI trained to maximize engagement. I am told that one could have used AI to promote other social goals, such as social harmony and cohesion.
That raises the question: Why have the social media companies—and other technology companies—used AI to pursue goals so antithetical to social welfare? A simple reason: Their objective is to maximize profits, not societal well-being. There are some limited circumstances where the two are coincident, where the pursuit of self-interest leads to societal well-being (as Adam Smith is widely interpreted as having suggested), but in areas like information and innovation, pursuit of profits can lead to harmful social outcomes—just as it did in the financial sector, where it became so evident that greed was not good.
Angwin: How do we turn the tide?
Stiglitz: We need a new economic model, and this is especially true when it comes to AI. I can’t describe it fully here, but besides the regulations to enhance competition and to prevent digital harms, we need more public investment. Today, we need more research on AI to be in the public domain. We also need to actively steer AI toward social objectives—rather than just letting markets drive it on their own. Unfortunately, policy has been steering innovation the wrong way, encouraging labor-replacing AI.
More broadly, we need a range of policies to deal with AI’s potential adverse consequences—we need to try to make sure that AI lives up to its potential of creating more shared prosperity.
As always, thanks for reading.
Additional Hello World research by Eve Zelickson.
P.S. There is no newsletter next week due to the Fourth of July holiday. I’ll be back in your inboxes on July 9.