The Future of Work in the Age of AI: A Debate Among Experts

Artificial intelligence is reshaping the world of work, raising urgent questions about productivity, inequality, and human agency. While its immediate effects are most visible in the U.S., China, and other advanced economies, AI’s disruptive potential is global.

At the CUNY Graduate Center, four leading thinkers—Paul Krugman, Daron Acemoglu, Zeynep Tufekci, and Danielle Li—joined labour journalist Steven Greenhouse for a probing discussion on AI and employment. Their conversation revealed both the promise and peril of our technological moment.

Choosing the Right Direction

Acemoglu framed the stakes clearly: AI’s future depends on the direction we choose. He criticised the current push toward Artificial General Intelligence (AGI), which he sees as unrealistic and job-eliminating. Instead, he called for “pro-worker AI” that enhances human capabilities rather than replacing them.

Krugman posed a deceptively simple question: “What is it good for?” He argued that AI won’t follow the usual pattern of benefiting the better educated. It’s not a straightforward labor-versus-capital story, but one of selective displacement across sectors.

Tufekci emphasised that generative AI breaks a core societal assumption—that tasks like writing essays or cover letters are inherently human. This shift undermines traditional markers of effort and merit.

Li warned of a structural shift toward a “superstar economy,” where a few individuals or firms dominate due to AI’s scalability. While AI models learn from human examples, they also transcend human limitations, raising concerns about the concentration of power.

Navigating the Near Future

Acemoglu cautioned that while foundation models are advancing rapidly, they thrive in controlled environments. Real-world jobs involve complex, interdependent tasks that AI cannot yet perform. He predicted a painful adjustment period over the next five years, followed by uncertainty.

Krugman expressed skepticism about the flood of investment in data centers, comparing it to the dot-com bubble. Unlike past tech booms funded by retained earnings, today’s AI infrastructure is often financed by debt—a troubling sign.

Li stressed that AI’s impact on occupations depends on implementation choices. Automating some tasks may degrade the overall job, making it less appealing. She predicted a shift in skill demand: analytical skills may decline in value, while social and emotional skills become more important.

Tufekci drew a line between utilitarian AI (e.g., coding tools) and engagement-driven AI designed to mimic human conversation. The latter, she warned, could manipulate users and distort expectations.

Inequality and Business Models

Acemoglu offered a compelling vision of AI as a tool for amplifying human expertise—imagine a system trained on the best electricians helping others improve. But he lamented that such applications lack viable business models. Most investment flows into digital marketing, not pro-worker tools.

Krugman noted that even highly paid, highly skilled jobs are being replaced, while others—like plumbing—remain untouched. He cited Kurt Vonnegut’s Player Piano to remind us that fears of machine displacement are not new. During the first 40 years of the Industrial Revolution, real wages stagnated despite technological progress.

Advice for Students and Workers

Li observed widespread FOMO among companies rushing to adopt AI. But productivity gains won’t come from buying tools alone—they require thoughtful integration and unique use cases. Her advice to students: build durable skills, embrace discomfort, and stay open to changing career paths.

Tufekci warned that AI’s ability to write human-like cover letters could push companies back toward elitist hiring practices—favouring referrals or top-tier colleges. Krugman added that some professors now require handwritten assignments to avoid AI-generated work.

Acemoglu challenged the comforting myth that technology always creates new jobs. Many weavers never recovered from the Industrial Revolution. AI won’t cause mass unemployment, but it will lead to real job losses unless we invest in new goods and services.

Krugman argued that the U.S. must increase research funding and support higher education to stay competitive. He contrasted America’s obsession with breakthroughs to his impression of China’s focus on doing smaller things well.

The Data Dilemma

Li raised concerns about how AI systems learn from workers’ data. Surveillance tools used for quality assurance are now feeding chatbot models that replace the very workers who generated the data. Companies no longer just rent labor—they extract and monetise behavioural data without consent or compensation.

She warned that while it’s tempting to admire a model that works while you sleep, it’s dangerous if that model replaces you and you’re fired.

Policy and Regulation

Tufekci proposed restricting AI systems from using human pronouns, arguing that conversational AI can exploit users psychologically. Krugman called for stronger regulation, rejecting the “move fast and break things” ethos when human lives are at stake.

Acemoglu advocated for forward-looking policy: taxing digital ads, reforming the tax code to favour labour, and establishing data property rights. Li urged policymakers to focus on low-hanging fruit—AI applications in drug discovery and healthcare that offer clear public benefit.

Note: Gen AI tool has been used to write this article.