AI pay gains could peak soon, but a Penn professor sees a way out (www.businessinsider.com)

University of Pennsylvania professor Ioana Marinescu — with coauthor Konrad Kording in a new Brookings paper — warns that AI-driven wage gains may be nearing a peak under what they call “intelligence saturation.” Using a task-based model, they find wages follow a hump-shaped path: initially rise as AI boosts productivity, then flatten and fall once automation reaches a critical mass. Their baseline simulation shows wage pressure setting in when about 37% of “intelligence” tasks are automated; Marinescu estimates roughly 14% are already automated now (drawing on long-term declines in routine cognitive jobs), and early signs of displacement show a 13% employment drop for 22–25-year-olds in AI-exposed occupations per a recent Stanford study. The paper’s key technical implication is that outcomes hinge on whether AI and human labor are complements or substitutes. If firms pair automation with real-world investment in physical capital (factories, hospitals, infrastructure), AI can amplify human productivity and sustain wage growth; if AI substitutes for human effort broadly, wages could decline. To avoid a downturn, the authors recommend slowing automation adoption, boosting investment in embodied physical work, and even taxing virtual labor to maintain a healthy machine-human balance — policy levers aimed at preserving complementarity and preventing broad labor-market reallocation toward lower-paid physical tasks.