Is the rapid adoption of artificial intelligence a job-killing guillotine or the greatest engine for professional reinvention since the industrial revolution? The tech industry loves to frame the conversation around the "Eureka!" moments of lone geniuses building chatbots in garages, but the real story here isn't the software itself — it’s the deliberate, often state-backed demand for the specific types of labor that software requires.
According to the MIT report, we have spent decades misunderstanding how new work actually comes to be. A study led by MIT labor economist David Autor and his colleagues, including Caroline Chin, Anna M. Salomons, and Bryan Seegmiller, suggests that innovation isn't just an accidental byproduct of a clever algorithm. Instead, it is a "purposive activity," often triggered by massive, concentrated investments like the government-led manufacturing expansion of the 1940s.
Think of it like building a new bridge. You don’t just need the engineers who designed the blueprints; you suddenly need specialized inspectors, high-altitude welders, and concrete chemists who didn't exist in the local labor market a year prior. When the government or the market dumps massive resources into a new frontier, they don’t just build products; they build entire career categories.
The findings, forthcoming in the Annual Review of Economics under the title “What Makes New Work Different from More Work?”, offer a sobering look at who actually benefits from these shifts. The data is clear: new work disproportionately favors college graduates under 30. While we often view tech as a democratizing force, the early stages of a new industry act more like an exclusive club. Because new expertise is scarce, it commands a wage premium. But like a shiny new piece of software that eventually becomes a standard utility, that scarcity eventually erodes.
The researchers, who utilized U.S. Census Bureau data spanning from 1940 to 2023, found that in 1950, about 7 percent of employees held jobs in fields that had emerged only since 1930. By the 2011–2023 period, that figure had risen to 18 percent for roles introduced since 1970. This transition follows a predictable lifecycle: a skill becomes specialized, it pays handsomely, it becomes common knowledge, and then it is automated. Remember when knowing how to use Microsoft Word was a resume-defining "specialty"? Today, that same skill is considered the baseline of professional literacy, as fundamental as knowing how to use a stapler.
This lifecycle is exactly what makes the current obsession with AI so volatile. We are currently in the "scarcity" phase where anyone who can prompt an LLM effectively feels like a wizard. But as Autor notes, if everyone is an expert, no one is an expert. The critical question for the average worker isn't whether AI will exist, but whether it will be used to hollow out existing roles or to augment the expertise of human workers.
Take the healthcare sector, which is heavily funded by public dollars. We have the leverage to force a market toward "socially beneficial" implementation—using AI to help workers with varying levels of expertise perform better, rather than simply replacing them. Whether we choose to use AI to build new, complex, and high-value roles or simply to strip the "human" out of the job depends less on the code and more on where we decide to place our institutional bets.
The next reading of the American Community Survey data will show whether the current wave of AI-driven job creation is genuinely expanding the workforce or simply accelerating the "scarcity-to-automation" cycle for a shrinking elite.
