Artificial intelligence, a force increasingly intertwined with every facet of modern industry and knowledge work, appears to be propelling a remarkable rise in productivity and wages. Yet, despite this moment of economic expansion, some scholars warn that the upward trajectory may not be sustainable indefinitely. According to Ioana Marinescu—an associate professor at the University of Pennsylvania’s School of Social Policy & Practice and coauthor of a new report from the Brookings Institution—the period of accelerating growth could eventually give way to a slowdown, a phenomenon she and her colleague Konrad Kording describe as “intelligence saturation.” Their analysis introduces a nuanced model of how automation reshapes the labor landscape, suggesting that the relationship between technological progress and wage growth is neither linear nor everlasting.
In their model, as automation spreads more deeply across sectors, wages are expected to trace what they term a hump-shaped trajectory: initially increasing as technology enhances worker productivity and business output, then stabilizing, and ultimately declining when machines begin to replace human cognitive labor rather than complement it. This dynamic implies that the very tools driving wage advances today could, if unchecked, contribute to stagnation or even erosion of income tomorrow. However, Marinescu emphasizes that this downward arc is not preordained. With deliberate policies and strategic investments, it may be possible to steer the economy toward ongoing shared prosperity. She cautions, however, that the present moment may mark an inflection point—perhaps the beginning of a phase in which artificial intelligence starts subtly suppressing wage gains.
To understand where we currently stand on this curve, Marinescu provides empirical estimates grounded in decades of labor-market research. She calculates that approximately 14 percent of what she defines as “intelligence tasks”—the cognitive, analytical, and decision-making processes once dominated by human minds—are now automated. This figure builds on studies showing that the share of routine cognitive jobs has dropped dramatically, from 49 percent in the late twentieth century to just 35 percent by 2018. This decline, she notes, positions the economy closer to the downturn phase of the wage curve than to its beginning. In the baseline Brookings simulation, wage contraction begins when about 37 percent of intelligence work is performed by machines, a milestone that could arrive surprisingly soon if current rates of AI adoption continue to accelerate.
As yet, Marinescu stresses, there is no definitive evidence that the overall wage landscape has begun to deteriorate. Observations remain preliminary. “It’s too early to tell,” she says, acknowledging that certain indicators are merely suggestive rather than conclusive. Nonetheless, she highlights emerging signals of disruption—most notably the displacement of less-experienced workers in occupations heavily exposed to AI technologies. Drawing on recent research from Stanford University, Marinescu points to early-career workers aged 22 to 25 in fields such as software development and customer support who have experienced roughly a 13 percent decline in employment since the advent of generative AI tools. By contrast, older or less AI-exposed employees have seen stable or even slightly rising employment levels. The broader structural warning, she argues, would appear when the overall proportion of intelligence-oriented jobs visibly contracts, signaling a deeper labor market shift toward more physically intensive work.
Still, a meaningful wage downturn is far from inevitable. According to Marinescu, the decisive factor will be how skillfully society orchestrates the transition between two interdependent domains: the “physical sector,” encompassing all work grounded in tangible production and real-world activity, and the “intelligence sector,” characterized by data, computation, and cognitive labor. In her view, these are not competing arenas but complementary forces—akin to the relationship between labor and capital, each essential for balanced economic growth. When automation evolves in complementarity with human skills, productivity continues to climb, and wage gains can persist. However, the paper also warns that the economic returns from continuously adding AI technology eventually reach a saturation point unless accompanied by comparable investment in physical capital and human capacity.
Translated into simpler economic terms, Marinescu argues that artificial intelligence and human labor can enhance one another’s performance, driving mutual productivity gains—provided that technological adoption does not outpace investment in sectors where human presence remains indispensable. The Brookings report underscores that the physical infrastructure of the economy—factories, hospitals, construction projects, and transportation networks—must keep advancing in parallel with intellectual automation. If innovation in the intelligence sector races ahead while the material economy stagnates, overall wage stability could falter. Hence, the researchers propose several strategies for preventing a collapse. Among them are moderating the rate of automation, channeling greater capital into physical industries, and designing fiscal policies that recognize the interdependence of human and machine labor.
One of the paper’s more provocative recommendations involves adjusting tax structures to account for “virtual labor,” effectively incentivizing firms to maintain a healthy equilibrium between human workers and AI systems. This idea echoes political proposals such as Senator Bernie Sanders’ “robot tax,” intended to ensure that companies adopting automation still contribute to social and economic well-being by supporting displaced workers or reinvesting savings into human employment. The underlying rationale is that a balanced distribution of technological benefits can mitigate inequality and foster sustained prosperity, rather than concentrating profits solely among machine owners or technology developers.
Ultimately, Marinescu emphasizes that the greatest uncertainty lies in the evolving degree of substitutability between AI and human output. Should artificial intelligence become capable of performing most cognitive tasks independently, wage growth could plateau once again, reflecting an economy where human contributions hold diminishing marginal value. Conversely, if the relationship remains complementary—each side amplifying the other’s strengths—then productivity, innovation, and worker compensation can continue to rise together. The challenge, therefore, is not only technological but profoundly social: determining how to integrate intelligent machines into the fabric of the economy in ways that preserve human significance and distribute progress equitably across society.
Sourse: https://www.businessinsider.com/ai-wage-gains-could-peak-soon-professor-on-solutions-investment-2025-11
