ChatGPT's threat to white-collar jobs, cognitive automation

ChatGPT and other large language models may put the value of many white-collar jobs up for debate. Some experts theorize that the large language model technology may bring about cognitive automation or task automation, such as code development and report writing, now mostly done by humans.

Anton Korinek, a professor in the Department of Economics and at the Darden School of Business of the University of Virginia, said in the next five or 10 years, he sees a diminishing role for humans in many cognitive tasks.

"We may increasingly turn into rubber stampers with a human veneer," he said. In other words, human workers merely approve a machine's work rather than contribute to completing a task.

Cognitive workers have jobs that require critical thinking and problem-solving skills. Cognitive automation involves automating the jobs now done by these typically white-collar workers. 

Korinek was among the participants at a forum hosted this week by the Brookings Institution and Georgetown University about ChatGPT and the future of work. The participants discussed the near- and long-term implications of this tool, the latest version of which launched Tuesday. But it needs to be clarified how consequential the technology will become.

There's a lot of anxiety among economists and the population about what [a large language model] means for the labor market and the future of work.

"We have been automating tasks and jobs ever since the Industrial Revolution," Korinek said. But what's different is that much of the automation has been around machines or blue-collar work and not cognitive workers, he said.

"There's a lot of anxiety among economists and the population about what [a large language model] means for the labor market and the future of work," said Sanjay Patnaik, director of the Center on Regulations and Markets at Brookings, at the forum.

It wasn't that long ago when "it looked like automation in blue-collar jobs, like self-driving cars and trucks, would accelerate very rapidly," Patnaik said. "The large language models have overtaken those advances in automation."

The debate over capability

What white-collar tasks could cognitive automation take over in HR? When queried, ChatGPT suggested the large language model could create personalized onboarding material and assist HR professionals in drafting documents, among other tasks.

There is still considerable debate about the capability of these tools. 

Susan Athey, a professor of the economics of technology at Stanford University's graduate school of business and a panelist, said the model is using pattern recognition, but "it's still not smart," she said. "The mistakes it makes also were predictable. Like if it learns from Reddit chats, it's gonna sound like a Reddit chat."

But Athey sees ChatGPT speeding up repetitive and frustrating research tasks. "The ability of ChatGPT to summarize information and not show you redundant information, I think, just supercharges any kind of research process," she said. 

Another panelist, David Autor, a professor of economics at MIT, said that many tools "make our skills and expertise and knowledge and creativity more valuable." But other tools can do the opposite and commoditize human skills and make them less valuable, such as a cab driver who has memorized maps only to see that skill commoditized by GPS, he said. 

People need to think about what they want to get out of the AI, Autor said. 

"We have a shared interest in directing the technology in a way that will be complementary to us, and therefore, advancing societal goals, helping us solve some of our hardest problems like climate change," he said.

In a separate interview, Mikaela Pisani, chief data scientist at Rootstrap in Los Angeles, said she sees ChatGPT as a useful technology that can help with a first draft that humans can then work on improving.

"We have to use it as a tool and not just replace what we do," she said. 

Patrick Thibodeau covers HCM and ERP technologies for TechTarget Editorial. He's worked for more than two decades as an enterprise IT reporter.