They say history never repeats itself, but it does often rhyme. It's why top researchers are looking to the past as a guide to predict how generative AI could affect workers' jobs in the years and decades to come. 

We asked four of these experts to unpack how new technologies and pre-ChatGPT AI have impacted jobs in the past, what this means for the workers of the future, and how you can prepare for the AI boom. 

Historically, technology-driven automation has displaced some jobs, but has also created better jobs and increased the volume of jobs in the long run, said Ethan Mollick, an associate professor of entrepreneurship and innovation at the University of Pennsylvania's Wharton School who requires his student to use ChatGPT. 

The big question is whether the new age of AI will have a similar story. 

"It's possible that in the end, we get better jobs, but in the short term, there's a lot of disruption," Mollick said. "We don't have a very clear model. This is the fastest-adopted general-purpose technology we've ever seen, and we don't know what its capabilities are ultimately."

One thing that everyone seems to agree on is that AI will increase the overall productivity of workers. That's expected to help buoy the economy and markets going forward. But Carl Benedikt Frey — an Oxford economist who coauthored a highly publicized 2013 paper that estimated 47% of all US jobs were at risk of being replaced by automation as soon as the 2020s — thinks there's a fine line between AI helping workers and AI hurting them.

"History tells us that simplification is often merely a step towards automation," he said, adding that, "AI assistants that analyze telemarketers' calls and provide recommendations are being trained with the ultimate goal of replacing them."

For historical reference, Frey pointed to the profession of lamplighters, the 19th-century workers who carried heavy torches and ladders to ignite the gas lamps that lit the streets at night. When electric streetlights started to take hold in the US in the late 1800s, the lampposts still had switches requiring human operation, Frey said. But eventually, substations started to control the streetlights, automating lamplighters' jobs at a large scale.

A different historical example could play out in many industries, Lindsey Raymond — a Ph.D. candidate at the MIT Sloan School of Management who previously worked as an economist at the White House — told Insider. She pointed to the cotton gin's invention in the late 1700s.

"The cotton gin made people who made clothes or cotton for clothes way more productive," she said. "But the price fell so much that there was this massive demand shift in the amount of cotton that people bought. So employment shot up."

While it is an apt historical example, we must also acknowledge that, at that time in American history, employers filled many of the industry's vacancies with enslaved people. 

Raymond pointed to a profession that's already faced AI disruptions for years — customer-service representatives — as a potential example of a similar scenario in action. 

"As part of COVID and this shift towards buying more things online, most companies have seen huge increases in the amount of demand for their online customer-support options," she said. "So with that occurring, I would not really expect to see negative employment effects.

But Raymond warned that AI could produce some less-desirable outcomes for customer-service workers, particularly if customer-support chatbots become much more capable and advanced. The productivity benefits that created jobs could also lead to increased competition and lower wages. What's more, the same scenario could plausibly unfold across other professions.

An April working paper from the National Bureau of Economic Research offered data supporting this idea: The newest and worst-performing customer-service workers in their study saw the biggest productivity boost from AI assistance. Put simply, it closed the gap between experienced employees and their potential replacements. This, in turn, could "bring down average wages quite substantially," Raymond said.

Oded Netzer, a Columbia business-school professor, said he still sees a way for more experienced customer-service workers to stand out in the jobs that remain.

"As AI improves, I do expect AI to replace some of call-center tasks, where the humans left in these jobs will need to be domain experts and deal with more difficult issues," he said.

The extent to which AI displaces jobs will depend on how quickly it scales what Mollick calls the "three levels" of work: tasks, jobs, and systems. Even once AI becomes adept at so many work tasks that it can arguably replace a job, the final "systems" hurdle remains. 

"Let's say AI gets better at diagnosing than your doctor in the short term, or AI is better at teaching a class than I am," he said. "There's still going to be a long time for change because systems take a long time to change. The students expect to see a human in the class, not an AI."

He said studies have found that the "most creative, highest-paying, and most-educated jobs," are the most likely to be transformed — not necessarily displaced — by AI, and that the closest thing to an AI "replacement story" is in the translation industry, where many human translators are still in the game but could face additional disruptions in the years ahead. 

Mollick said the late-19th century automation of telephone operators — then a common profession for women — suggests that when job displacement does happen, older workers could struggle. 

"When you got rid of operators, then basically young women were able to adjust, find new jobs and were able to adapt," he said. "But older women took a lifelong hit to wages — were never able to find as good a job again."

Mollick has one piece of advice for workers looking to adapt to the AI boom in the years ahead: Rather than focusing on the tasks AI can't do, learn how to use it to make yourself more productive. 

"Everyone expects these models to improve in power," he said. "So you can't necessarily make a bet that AI is bad at this now and will always be bad. Instead, what I would be thinking about is: How do you figure out how to use it to do your job better?"