TEcologists tend to predict that the economic impact of their creations will be unprecedented. This is especially true when it comes to artificial intelligence. Last year, Elon Musk predicted that continued advances in AI would make human labor obsolete. OpenAI CEO Sam Altman wrote that AI will inevitably continue the shift of economic power from labor to capital, creating “tremendous wealth.” Jensen Huang, CEO of semiconductor design company Nvidia, likened the development and deployment of AI to “a new industrial revolution.”
But while technologists are bullish about the economic impact of AI, economists, another member of the tech clergy that has a huge impact on people's lives, are not. Even if engineers develop the powerful AI systems they say are on the horizon, the economic impact of those systems is likely to be overwhelming, many economists say. According to George Mason University economics professor Tyler Cowen, AI “will boost annual growth rates in the United States by one-quarter to one-half of a percentage point.” Nobel laureate in economics Paul Krugman has argued that “history suggests that it will take longer than many currently expect for the significant economic benefits of AI to be realized.'' . “The industrialized world is full of jobs, and will continue to be so,” writes David Autor, an economics professor at the Massachusetts Institute of Technology.
Anton Korinek, a professor of economics at the University of Virginia and affiliated with some of the most respected institutions in the profession, is a certified member of the priesthood. However, the 46-year-old broke ranks. He is using techniques from his own field to model what would happen if his AI developed as many engineers say. This means that AI systems will likely outperform humans at every task by the end of the decade. Within the barren vocabulary of academic economics, he paints a disturbing picture of a near future in which humans play no role in the economy and inequality increases.
“If the complexity of tasks that humans can perform is limited and full automation is achieved, wages will collapse,” Korinek and one of his PhD students write. student Donghyun Suh said in his recent paper. In other words, if there is a limit to the number of complex tasks that humans can perform, and machines become advanced enough to fully automate all those tasks, then human labor will no longer be needed, wages will drop significantly, and people who do not share will You may be left behind. This could lead to resource starvation for AI systems and their owners.
Economists have spent the past 200 years explaining to the uninitiated that the lump-of-labor fallacy, the idea that there are a fixed number of jobs and that automating those jobs will create permanent unemployment, is false. Korinek said. “In a way, it’s our professional baggage,” he says. “We've spent so much time fighting false narratives that it's hard to pivot when the facts actually change and see that this situation could actually be different. ”
“The most frightening thing about this is that the technology predictions that I'm using as input are those of Sam Altman and [Anthropic CEO] Dario Amodei is free to preach in public wherever he is,” says Korinek. “I'm not a technology expert, so I don't have any special insight or knowledge into these issues. I take what they're saying seriously and what it actually means for the economy, jobs and wages. I'm just asking if it would be completely destructive.”
Korinek's father, a primary care physician, instilled in her a passion for understanding the brain. During his teenage years living in Austria, Korinek became fascinated with neuroscience and learned programming. In college, he took a course on neural networks. Neural Networks is his brain-inspired AI system that has since become the most popular and powerful in the industry. “It was interesting, but also boring, because even with the tools we had at the time, the technology wasn't very powerful,” he recalls.
Korinek was about to graduate from college in the late 1990s when the Asian Financial Crisis occurred. His 1997 crash caused by the Thai foreign exchange crisis was followed by a number of episodes of severe financial crises in emerging markets. Korinek decided to pursue his Ph.D. About financial crises and how to prevent them. Fortunately for him, and not so fortunately for others, the global financial crisis occurred in 2008, the year after he completed his doctoral thesis. His expertise was in demand and his career took off.
As an academic economist, he watched from the sidelines as AI researchers made breakthrough after breakthrough. He reads newspapers, listens to radio programs, and reads books about AI. In the early 2010s, as AI systems began to outperform humans at tasks such as image recognition, he believed that artificial general intelligence (AGI) was a term used to describe not-yet-built AI systems that could perform any human task. started. It could be developed surprisingly soon. However, it was the birth of his first child in 2015 that brought him off the sidelines. “Something is really happening, and it's happening faster than I thought in the 1990s,” he recalls thinking. “This is going to be really relevant to my daughter's life.”
read more: Four charts showing why AI progress is unlikely to slow down
Korinek uses economics methods to explore how future AI systems may impact economic growth, wages, and employment, and how the inequalities created by AI may undermine democracy. We have sought to understand the potential risks and how policymakers should respond to the economic challenges posed by AI. For many years, addressing the economics of AGI has been a fringe pursuit. The economics doctoral students told Korinek that they wanted to work on the problems he had begun researching, but that he felt he needed to stick to more mainstream topics if he wanted to get the job. Even though he won his tenure in 2018 and was free from the pressures of the university's job market, Korinek felt discouraged at times. “Researchers are also still social creatures,” he says. “When he has so much skepticism about his work, he starts to question himself and it becomes very difficult for him to advance his research agenda.”
Until OpenAI released its wildly popular chatbot, ChatGPT, in November 2022, AGI's work on economics was a fringe field, Korinek said. Now, it's “on an exponential growth trajectory,” Korinek said.
Many economists still, quite reasonably, deny the possibility of AGI being developed in the coming decades. Skeptics point to potential obstacles to continued progress in AI, including the many ways current systems fall short of human capabilities, lack of data to train larger models, and concerns that engineers may be overconfident. It points out the history of making predictions. A paper published in February by economists from Portugal and Germany brushes aside AGI as “science fiction” and argues that AI is unlikely to spark explosive economic growth.
read more: When will AI outsmart us? It depends on who you ask.
“Two years ago, that was completely normal. That was the median response,” Korinek said. However, more and more people are becoming interested in the possibility of AGI being developed, and those who continue to deny the possibility are now changing their position rather than denying it outright. Provides a solid discussion. “In other words, people are taking the discussion very seriously,” he says approvingly.
Other economists are positive that AGI will be developed in the foreseeable future, but argue that it will not cause a job collapse. Often, they attribute their disagreements with those who think the development of AGI could cause these things to economic ignorance on the other side.
For example, in March, economist Noah Smith made the claim in a blog post that was picked up by a New York paper. times Even if AI were to outperform humans at any job, there would still be many high-paying jobs. Smith patiently explains that this is because AI systems and humans can specialize and trade with each other for tasks where each has the greatest relative advantage (an economic concept known as comparative advantage). “When most people first hear the term 'comparative advantage,' they immediately think of the wrong thing,” Smith writes.
But Korinek is not economically ignorant. “I claim to understand comparative advantage,” he says with a laugh. Korinkek's model, which takes into account comparative advantage, still does not paint a rosy picture for humans in a world where AI systems can perform every task that humans can. “The theory of comparative advantage shows that there is scope for gains from trade,” Korinek says. “But the big problem is that it doesn't tell us whether the terms of that trade are sufficient for humans to make a decent living, or for humans to cover their subsistence costs.”
Later in his blog post, Smith said that in order to ensure that humans earn enough to survive, governments need to be able to raise taxes on AI and place limits on the amount of energy that AI systems can use. It goes on to say that intervention may be necessary. If only AI could outperform them at every task. But while Smith is confident the government will take such steps if proven necessary, Korinek has less confidence in him.
“Economists are gradually starting to take seriously the potential for AI to destroy jobs,” he says. “Policymakers aren't there yet — that's my concern. I'm nervous because we may not have much time.”
Korinek does not share the confidence that many technologists predict about AGI's imminent development. But given the potential impact of such technology, society needs to be prepared, he says.
“It may take many more decades before we even come close to replicating what the human brain does,” he says. “But I also think there's a good chance that it won't actually take that long and that this will happen really quickly and upend our society.”