Make Artificial Intelligence Pay
Adding new meaning to the term “fake news,” a recent Congressional hearing opened with a voiceover of a sitting U.S. Senator recorded using a voice mimicked by artificial intelligence (AI) and a script written by ChatGPT. If one were simply listening to the hearing, no one could tell the difference between the fake or real voice. This invites the question whether legislators and their staff will be needed in a future dominated by AI.
Many are rightfully concerned about the potential threat that AI and large language models (LLM) pose to the economic security of current and future workers. In Megathreats, economist Nouriel Roubini writes: “The possibility is very real that a tiny top echelon will win while everybody else loses their jobs, their incomes, and their dignity.”
According to the Brookings Institution, approximately 25 percent of U.S. employment, the equivalent of 36 million jobs, will have experienced high exposure to automation of 70 percent or higher by 2030. If we are moving toward a world where an increasing percentage of work is performed by robots, while personal data continues to be extracted and monetized by profit-making entities, then the question of how to support the viability of human enterprise is paramount.
Some have promoted the concept of paying people for their personal data, also called “data dignity” or “data as labor,” as an alternative to the current system in which people give away their data in exchange for free internet and app-based services.
Another approach to be seriously considered is taxing AI technology in a way that supports human endeavor and dignity without stifling innovation. Workers displaced by AI will need to have their lost or reduced earnings replaced. There are viable ways to generate substantial revenues from AI to support social insurance programs, like Social Security, that replace wages and provide economic security for human workers and their families over a lifetime.
Social Security’s financial sustainability depends greatly on wage-based contributions by employees and employers. But, if human wages decrease due to increased use of automated technologies in the workplace, then the finances of Social Security’s Trust Funds — which are currently projected to experience a shortfall of 25 cents on every dollar of promised benefits within the next ten years or so — will continue to be destabilized even if Congress addresses the shortfall.
To counteract this effect, Congress should create the Federal Artificial Intelligence Contributions Act (FAICA) as a complement to the Federal Insurance Contributions Act (FICA), which is the current way Social Security’s retirement, disability, and survivor benefits are largely funded. It is not only fair but also necessary that we subject work performed by non-humans to similar fiscal treatment levied on human labor.
Imagine ways in which the work performed by non-human entities might be valued for the purpose of applying FAICA. One possibility would be to impute to each robot the lost wages of workers whose labor has been disrupted. Another would be to place specific, time- adjusted dollar values on AI productivity, perhaps using an AI tool to quantify value over time. FAICA revenues to the Social Security Trust Funds could be transferred through the same mechanism employers currently use for FICA, or through newer technology, such as blockchain or central bank digital currency.
Since robots won’t be retiring and claiming Social Security for themselves, revenues generated by FAICA could be used to expand Social Security. It could also be used to add a basic income floor and/or a baby bond benefit to Social Security.
Revenues could also be decoupled from current Social Security benefits entirely to finance a new equity-based U.S. sovereign wealth fund, which could generate additional revenue by investing in stocks, start-ups, and other investments beyond U.S. Treasury securities, which by law is the only investment Social Security is currently allowed. Additional revenue generated by this approach could be used to create or supplement all the aforementioned methods for ensuring economic security for human workers and their families.
Some might reasonably challenge whether any of these approaches would be enough to help those persistently unemployed or living in poverty. This is a valid critique and necessary to address. AI is projected to have faster destructive effects on jobs traditionally held by lower- and moderate-income workers, who are disproportionately composed of teens and young adults, gender minorities, non-white racial and ethnic minorities, and persons who are differently abled. This means that benefits in some future AI- supported program structure must be creatively designed to protect these groups.
Traditionally, programs have sought to offset labor market disadvantages by making benefits “progressive,” which means providing an increasingly larger percentage of replacement income to those who experience longer spells of unemployment or lower lifetime earnings. But in a world where AI will likely eliminate starter jobs for young people and algorithmic bias could create labor market disadvantages for others, new approaches must be devised to fairly compensate those whose incomes suffer disproportionately.
AI is projected to add $15.7 trillion to the global economy by 2030. Such newfound global wealth offers an extraordinary opportunity to revolutionize public policy in an effort to reduce inequality and poverty in the U.S. and worldwide. The U.S. needs to lead the way in developing innovative policy solutions that turn AI-generated labor market instability into guaranteed economic security for all.
Dr. Maya Rockeymoore Cummings is a retirement security and health policy expert and a non-resident senior fellow at the Brookings Institution in Washington, DC.
William “Bill” Arnone is a retirement security, employee benefits expert, and the CEO of the National Academy of Social Insurance in Washington, DC.
