A decade ago, two academics at the University of Oxford published a paper that would be cited for years to come. It was called “The Future of Employment: How Susceptible Are Jobs to Computerisation?” and its headline answer to that question—47% of U.S. jobs—laid the foundation for the current trend of being rather scared of AI’s societal impacts.
Now, in the midst of a generative AI craze that few could have foreseen back in 2013, Carl Benedikt Frey and Michael Osborne have revisited their seminal paper, asking themselves whether they had underestimated automation’s near-term effects on employment. I had a look at their working paper, which will be published in January, and was struck by the persistence of the bottlenecks they previously identified. While genAI will kill off many jobs in virtual settings—telemarketers beware—Frey and Osborne write that these limitations will keep AI in the realm of relatively low-stakes tasks for a while yet:
“In-person interactions remain valuable, and such interactions cannot be readily substituted for: LLMs don’t have bodies. Indeed, in a world where AI excels in the virtual space, the art of performing in-person will be a particularly valuable skill across a host of managerial, professional and customer-facing occupations. People who can make their presence felt in a room, that have the capacity to forge relationships, to motivate, and to convince, are the people that will thrive in the age of AI. If AI writes your love letters, just like everybody else’s, you better do well when you meet on the first date.”
Add to that generative AI’s tendency to “hallucinate,” and companies are going to hold back from entrusting AI with anything involving longstanding relationships. And when it comes to the arts, genAI still doesn’t do originality. “It is no coincidence that AI does best in tasks where we know what we want to optimize for, like for the score in a video game,” the authors write. “Yet if the goal is to generate something entirely new, for what do you optimize?”
Put another way, generative AI will do a great job of churning out text in the style of Shakespeare, Frey told me over the phone today, but “Shakespeare already existed and, if you want to do something entirely new, what do you prompt it for?” The technology suits the creation of “sequels rather than breakthroughs,” he added.
Also, AI systems still work best in structured environments (like automated warehouses) rather than those that serve up surprises (like the open road).
Generally, Frey and Osborne see AI’s big role in the near future as one of simplifying in-person tasks rather than completely taking them over. This could make it easier for less-competent people to enter industries, which could increase competition and ultimately depress wages—like Uber did in the taxi industry—but it’s not necessarily going to kill that many jobs. Because of that, and the fact that genAI’s drawbacks are unlikely to be solved anytime soon, the Oxford academics are holding off from issuing a new headline figure for at-risk roles.
“We don’t think these bottlenecks will persist indefinitely, but we think they remain relevant today even in this new age of generative AI,” Frey said. “Generative AI is not yet an automation technology—it still requires a human in the loop for various reasons.”
More news below. And if you’re still in a genAI frame of mind, check out the new principles proposed by the U.K.’s Competition and Markets Authority for the technology’s further development.
Want to send thoughts or suggestions to Data Sheet? Drop a line here.
David Meyer
