Keep up with the innovative tech transforming business
Tech Brew keeps business leaders up-to-date on the latest innovations, automation advances, policy shifts, and more, so they can make informed decisions about tech.
The effect AI could have on particular jobs has been a key question of the generative AI era.
A team of researchers attempted to comprehensively answer this question by having human annotators rate tens of thousands of tasks across more than 1,000 occupations based on their exposure to large language models (LLMs). Their paper, published recently in Science, found very few examples of jobs that could be mostly performed by AI, though the authors said the number has the potential to balloon with time.
The team, which included researchers from the University of Pennsylvania’s Wharton School, Oxford, and OpenAI, concluded that broad generative AI could perform more than half the tasks of just 1.8% of jobs in its current state. However, when accounting for potential domain-specific software developments that could complement AI capabilities, the number could jump to as much as 46% in the future, the authors wrote.
White-collar danger? Unlike previous waves of automation technology that tended to displace lower-wage workers first, generative AI is more likely to affect higher-wage jobs with greater levels of preparation necessary, from lawyers and pharmacists to scientists and technologists, the study found. This is consistent with the findings of other studies that have shown office workers to be more at risk.
Jobs with more years of training, education, or experience required—“lawyers, pharmacists, database administrators”—were more exposed than those with lower barriers to entry—“dishwashers, floor sanders.” The two job clusters most exposed to risks of future job loss from generative AI overall were “scientists and researchers,” then “technologists,” which includes software engineers and data scientists.
The good news for workers is that it will likely take further advancements and additional software for this threat to be fully realized.
To augment or replace? While proponents of AI are fond of claiming that the technology will supplement rather than replace existing workers, the authors counter that it’s trickier to suss out the ultimate upshot of this tech—jobs that are highly exposed are perhaps as likely to be replaced as they are to be augmented.
“This challenges the notion that it may be possible to predetermine whether LLMs will augment or automate jobs, despite recommendations for AI to complement rather than replace human work,” the authors wrote in the study.
The authors said the findings point to the need for more policy discussions about how to best mitigate the impact of AI on jobs.
“Our research highlights the need for robust societal evaluations and policy measures to address potential effects of LLMs and complementary technologies on labor markets,” they wrote.