- Since the 1980s, automation has replaced middle-income workers with robots and software.
- Automation has also made educated labor more valuable and contributed to income inequality.
- Artificial intelligence can reduce educated labor and reduce inequality.
Automation has widened income inequality in the United States. Recent developments in artificial intelligence may reflect this.
In research released in March, OpenAI, the company that created ChatGPT, found that the latest AI could affect white-collar workers more than the rest of the workforce. Technology could dampen the explosive wage growth they experienced at the expense of their blue-collar counterparts by shifting more tasks to middle-skilled workers, which could reduce income inequality. But this rosy future depends on how AI evolves and whether we choose to regulate it.
Up until this point, automation has mostly hurt blue-collar workers. Daron Acemoglu, an economist at the Massachusetts Institute of Technology, found that automation was a driver 50% to 70% of the increase in wage inequality in the United States since the 1980s. Machines, robots, and computer software have replaced workers without a bachelor’s degree in routine tasks, such as assembly line work or administrative office work. Those workers who are less in demand have seen their real wages decline or stagnate, especially for those who previously made middle-class incomes.
This automation placed a premium on skilled white-collar labor and increased wage inequality. While automation has devalued and replaced lower-skilled work, the earnings premium between workers with a bachelor’s degree—a proxy for high skills—and those with a high school diploma multipliedFound David Autor, an economist at the Massachusetts Institute of Technology.
“Not only does it not replace you, but it makes your expertise more valuable,” Autor told Insider.
Replacing less skilled human labor with machines increases demand and wages for abstract tasks that cannot be easily automated. As a result, wage inequality has increased dramatically over the past half century.
Today, those same white-collar workers who benefited from inequality are now in the crosshairs of artificial intelligence.
“The general assumption is that AI — and even before that, office software — will affect mid-skill workers,” Acemoglu told Insider.
“We are really in a different era,” Autor said.
Whereas previous programs relied on written instructions to carry out routine tasks, such as a recipe that yields the same meal, the AI can be trained to come up with answers to never-before-seen data, allowing it to learn more challenging tasks like how to ride a bike.
He added that it is “powerful and opaque” because “computers now know more than we do”.
OpenAI march paper It found that big language models — the type of artificial intelligence behind ChatGPT, which can process and generate text that looks remarkably like natural human language — will impact at least 10% of tasks for 80% of the workforce and 50% of tasks for 19% of the workforce. , “with higher-income jobs likely to experience greater exposure to LLM capabilities and LLM-powered programs.” The report also found that holders of bachelor’s, master’s and professional degrees were more likely to be affected.
Exposure means that AI can replace high-income workers, reducing inequality, or make them more productive, which will increase it.
If technology is as capable as technologists hope it is, AI “can reduce inequality” and “be an equalizing force,” as it will replace highly skilled workers and reduce demand for their wages well, Acimoglu said.
Even if technology is not advanced enough to replace these highly skilled workers, it may increase the productivity and quality of medium-skilled work, and thus increase average income.
“You can imagine passing some of the most highly skilled tasks to less skilled people,” Autor said. For example, AI could allow nurse practitioners to “do what a doctor does traditionally because this tool is powered.”
Acemoglu said reducing inequality with AI is not “a crazy scenario, but there is a lot of uncertainty”.
“We still don’t know the full capabilities of large language paradigms,” he added.
“This may turn out to be a technology that widens inequality, rather than negatively affecting more skilled, higher-paid workers,” he said. Give two reasons.
First, “we tend to underestimate the importance of complex human tasks,” he said, so it’s possible that AI could supplement the work of journalists, managers, and radiologists, rather than replace them. This can increase their productivity and wages.
In addition, “more skilled workers may be able to protect themselves,” he said. If their job is automated, they can fill the next tier of higher paying jobs.
“They will not be affected as badly as they were when the robots were introduced,” he added.
“It’s a new game, and we don’t know what that new game will be,” Acemoglu said.
Autor said he believes AI “could reasonably contribute to reducing inequality” for those pushed out of their middle-income jobs, but that this would require changes in how workers are trained, employers are employed, and government is driving innovation “to make sure That companies don’t just automate things away.
Yet, he said, “incentives are not well aligned with social goals.”
“All of this is being played out in an arms race by a group of for-profit companies,” he added. “There is no reason to assume that what is good for Amazon and Google is what is good for humanity.”
However, he said, “the future is a choice” that is affected by the way we deploy our tools.