AI and the workplace

Think:Act Magazine "It’s time to rethink AI"
AI and the workplace

May 15, 2024

Artificial employees and workers need to get along

Listen to the article

Article

by Steffan Heuer
Artworks byCarsten Gueth

AI exists to empower and complement workers, not to fulfill engineers' wild dreams of surpassing the human mind. Hence, in the workplace, the focus is on machine usefulness.

Daron Acemoğlu, a labor economist and prolific author on the vagaries of wages, has a simple rule of thumb to judge AI: What's in it for us humans, particularly workers? According to him, the verdict so far is anything but encouraging. He laments that economies around the world are falling for what he calls the "AI illusion," or the misconception that new technology, including intelligent machines, will bestow unimaginable benefits upon us. "It's a continuation of the view going back to the 1950s that there is a great social value to making machines intelligent and autonomous, and that it's both a desirable and achievable aim," the professor at the Massachusetts Institute of Technology says in an interview.

Daron Acemoğlu, MIT professor and labor economist.
Daron Acemoğlu, MIT professor and labor economist.
"Instead of fixating on machine intelligence, we should ask how useful machines are to people."

Daron Acemoğlu

Professor and labor economist
MIT

Most if not all AI development, according to Acemoğlu, is focused on automating tasks, first the simple routine ones and now increasingly nonroutine or more complex tasks. The well-funded and highly competitive race to achieve machine intelligence that by some measure may surpass human capabilities is a dead end, he warns, because it falsely pits humans against silicon. Acemoğlu argues that society and businesses, however, should focus on a different metric called "machine usefulness" instead to gauge what AI can do for people with blue- and white-collar jobs: making their work more productive and meaningful. "Instead of fixating on machine intelligence, we should ask how useful machines are to people," he writes in his latest book Power and Progress: Our 1,000-Year Struggle Over Technology & Prosperity.

Machine usefulness is based upon the postulation that technology ought to be in service of people and complement them. Acemoğlu spells out four ways in which digital technology such as new AI systems can be steered toward this goal: by improving the productivity of workers in their current jobs; by creating new tasks with the help of machine intelligence augmenting human capabilities; by providing better and more usable information for human decision-making; and by building new platforms that bring together people with different skills and needs. All things that co-pilots or AI-powered programs can do – if we only gave up the conceit that tools like ChatGPT need to be as intelligent as us, Acemoğlu says.

Daron Acemoğlu

Daron Acemoğlu specializes in the interplay between labor economics, technological change, economic development, growth and inequality. The MIT economist has written or co-authored six books, including Why Nations Fail: The Origins of Power, Prosperity, and Poverty and Power and Progress.

The concept of putting workers' needs first is not that far-fetched. When fellow researchers at Stanford and MIT analyzed the effect of a new, generative AI tool on call center workers, they discovered that it boosts productivity by up to 30% because it "disseminates the best practices of more able workers and helps newer workers move down the experience curve." Giving workers a hand, in other words, can lift all boats instead of making most jobs sink. And long-term assessments seem to confirm this evaluation. In 2013 Carl Benedikt Frey and Michael A. Osborne, a pair of Oxford economists, published their seminal paper entitled The Future of Employment: How susceptible are jobs to computerisation? It went on to create a stir because it put almost half of all US jobs at risk of being eliminated. Yet when revisiting their findings 10 years on, the duo thinks that computers are nowhere near taking over: "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."

Maintaining this distinctly human advantage means learning new skills, however. When IBM surveyed executives in 2023, 40% expected their workforce will need to reskill in response to AI and automation, potentially impacting 1.4 billion of the 3.4 billion people in the global workforce, according to World Bank statistics.

For Acemoğlu, the current wave of excitement over artificial intelligence is nothing new, but rather another chapter in the long history of techno-optimism, a belief that tends to underestimate human skills and overestimate machine skills to ultimately serve the agenda of the entrepreneurs who stand to reap huge windfalls while sidelining the vast majority of people. "The last thousand years are filled with instances of new inventions that brought nothing like shared prosperity," he writes.

This more somber view of technological progress notwithstanding, Acemoğlu calls fears of automation wiping out millions of jobs overblown, at least for now. "Research shows it did take jobs, it did lead to wage losses and it did lead to greater inequality, but people found jobs in other places. Will we ever go to mass unemployment because of AI? Probably not in the next 30, 40 years," he says. What's overlooked, however, are the hidden costs in terms of rising wage inequality, eventually leading to a two-tiered society with a tech elite controlling vast datasets and the tools that leverage them.

WHY AI IS NOT YOUR CO-WORKER

Joanna Bryson wants us to be wary of the new office mates

Knowledge workers who enthusiastically embrace their new, synthetic "colleagues" should heed the advice of Joanna Bryson, professor of ethics and technology at the Hertie School in Berlin.

"AI is not your new friend or your new co-worker," she says, pointing to the fact that true collaboration is aligned around comparable intentions, moral agency and responsibility.

"AI is an extension of capital and management – something that your company has built or paid another company to build, and both express their goals through it. You're not working with the AI, but you're working for it. If AI makes your job more fun, that's great. But at the end of the day, you need to work with your human co-workers to make sure you're getting adequate pay and protection and everything else from your company. AI will not be your advocate."

Making sure that software serves employees and not the other way round requires redirecting technology's trajectory while there is still time, Acemoğlu says with a hint of hopefulness. "The machine usefulness path has become even more promising with generative AI tools, because many of us are in the business of problem-solving. But first, we need to change the narrative and identify the problem. We are having the wrong conversations today, whether we are the luckiest generation that has ever lived or whether killer robots will destroy us all."

The far better approach, according to the academic, is to develop and promote an alternative narrative around how AI should, first of all, benefit workers and make citizens better informed. In a second step, countervailing powers need to be built, from new institutions to new norms and regulations. While Acemoğlu admits that the power of labor unions and consumer protection organizations has waned, it is worth trying. "It's a hard thing, but not impossible. If we become more and more trusting of the tech geniuses who say they are going to save us, it becomes an inescapable trap. The point is not to oppose technology, but to point out how you can use it better."

The Al impact on the boardroom, the workplace, society and you. Please click below to read the other parts of the cover story:
About the author
Portrait of
Steffan Heuer has been covering the intersection of technology, commerce and culture in Silicon Valley for more than two decades. His work has appeared in The Economist, the MIT Technology Review and the German business monthly brand eins. He currently divides his time and reporting between Berlin and California.
Download the full edition
Think:Act Edition

It's time to rethink AI

{[downloads[language].preview]}

As our lives and work are increasingly being reshaped by AI, the new edition of Think:Act Magazine explores what this means for C-level managers.

Published May 2024. Available in
All online publications of this edition
Load More
Portrait of Think:Act Magazine

Think:Act Magazine

Munich Office, Central Europe