“We could be doing much, much more for the millions of Americans who are going to be displaced,” Yang told us in a recent interview, amid a flood of new figures that confirm the somber assessment, thanks to the joint efforts of researchers at MIT, the IMF, and McKinsey, showing the fragility of the U.S. job market’s future as a result of AI.
Yang’s warning is no longer the concern of a futurist but a fact. According to the most recent Iceberg Index released by MIT, current levels of AI technology can do work that comprises 11.7% of the U.S. job market, or roughly $1.2 trillion worth of work, in finance, healthcare administration, logistics, professional services, and other occupations. Unlike previous estimates that defined “exposure” in terms of future possibility, this indicator measures capability within the near term. Developed with the use of ORNL’s Frontier supercomputer, this model replicates 151 million employees as ‘agents,’ with ‘32,000 skills associated with different AI capabilities.’
Yang’s own metric, that 44% of U.S. jobs involve repetitive manual or cognitive tasks, lines up with other research. According to the IMF, 60% of employment in developed countries will be impacted by AI, with half of those being negative effects. McKinsey’s report indicates that some technologies theoretically may be able to automate more than half of the hours worked in the U.S. Starting with Yang’s own number of 44% of jobs, if half of those were automated in a decade, there would be as many as 30 to 40 million lost jobs, Yang estimated.
Evidence of acceleration is already apparent. Amazon’s own internal strategic plans, as reported by The New York Times, show a plan to automate 75% of their business, thereby eliminating the necessity for additional employee hiring, as estimated, of more than 600,000. Amazon’s Vulcan system enables robotics pick-and-stow functions within their warehouse shelves, although Amazon’s Robotics leadership claims this as a “collaborative” system, their course feels increasingly aligned with decreased human employee count in its fulfillment centers. Such models can be seen across other professional industries as well, with Salesforce, Walmart, HP, IBM, and Fiverr announcing layoffs linked to increasing AI adoption. IBM’s own Arvind Krishna, its CEO, admitted that they did not hire thousands of employees in its back-office functions that would be better automated via AI.
From a technological standpoint, the impact of AI goes well beyond chatbots and programming assistants. Large-scale language models and software agents are already capable of performing document processing, financial analysis, and HR administration. When it comes to logistics, self-directed systems control routes, inventory, and, through robotics, the physical transportation of goods as well. Such applications dissolve the barrier that protected mid-skill administrative and operational positions that were believed to be safe from AI.
Yang’s policy solution is a resurrection of his “Freedom Dividend,” a basic income guaranteed for every American adult of $1,000 per month, financed through those who benefit most from the scale-up of AI. Yang cites measures such as Dario Amodei’s “token tax” charge on AI corporations or a more comprehensive “compute tax” on the massive processing powers fueling AI models. Given that ‘tech giants were already worth “hundreds of billions,” Yang asserts that such taxes would yield “very big numbers very quickly.” At $12,000 per person per year, the dividend would be but a fraction of the approximate $85,000 GDP per capita that the U.S. generated in 2024.
The debate over UBI is increasingly being driven by the same AI executives responsible for the systems displacing workers. Elon Musk, for example, describes this as ‘just something that’s going to happen.’ But sociologists point to “symbolic violence,” with policies masquerading as well-intentioned but essentially codified as permits for greater control by AI’s owners. An OpenResearch project, led by Altman, found that although $1,000 monthly handouts supported the necessary, there was ‘no statistically significant improvement in job quality, education, or health.’
Engineering, economic models, as well as studies, such as that of David Autor, an economist from the Massachusetts Institute of Technology, confirm that this impact will be uneven. Meanwhile, skill change analyses carried out by McKinsey forecast that more than 70% of existing skills will remain relevant, although they will be exercised in a different manner, with a huge focus on AI literacy as well as other human skills such as negotiation, problem-solving, and analytical skills.
Policymakers and executives must face the technological truth: AI already offers cost advantages for a substantial portion of the workforce, with its adoption increasing with regard to both physical and cognitive capabilities. But the task, as difficult as it is to provide a safety net for the job losers, lies also in preparing a system that captures the value of this productivity shift, estimated to be in trillions, with its benefits shared, not merely enjoyed, by the AI elite.
