Nowadays there seems to be nonstop discussion about AI, with much of the conversation focused on whether there\u2019s a speculative bubble or whether the chipmaker Nvidia is really worth $5tn or whether OpenAI will beat its rivals in developing new generations of artificial intelligence. But the vast majority of Americans \u2013 just like the vast majority of Europeans and Asians \u2013 couldn\u2019t care less about those things.<\/p>\n
Their big concern is whether AI is going to cause huge layoffs and create a disastrous job market, especially for younger workers. Dario Amodei, CEO of Anthropic, a leading AI company, fed those fears when he said that AI could wipe out half of all entry-level white-collar jobs<\/a> in the next one to five years and increase unemployment in the US to 10% to 20%. In October, Bernie Sanders, the top Democrat on the Senate education and labor committee, issued a report saying AI and automation could replace up to 97m jobs<\/a> in the US over the next decade.<\/p>\n Such predictions fuel worries that AI will make today\u2019s enormous income inequality even worse as the already wealthy investors in AI grow even wealthier while millions of workers lose their jobs and perhaps form a new underclass struggling to make ends meet.<\/p>\n In a recent panel discussion<\/a> I moderated, Daron Acemoglu, an MIT economist and winner of the 2024 Nobel prize in economic sciences, said there were essentially two routes for developing AI: an anti-worker route and a pro-worker route. He voiced dismay that tech companies were focused on the anti-worker route \u2013 a route that aims to develop AI in ways that maximize automation and maximize job reductions.<\/p>\n In that panel discussion<\/a> at the City University Graduate Center in New York, Acemoglu said AI could take \u201cvery different directions, and which direction we choose is going to have great consequences in terms of its labor market impact\u201d. He said that today\u2019s AI \u201ccraze is really an automation agenda\u201d which \u201cis going to eliminate more and more jobs\u201d.<\/p>\n Acemoglu called for \u201ca different future\u201d with \u201cpro-worker AI\u201d. In his view, it would be far better if society and government could get tech companies to develop AI in ways that, instead of maximizing layoffs, increase workers\u2019 skills so that workers become more capable and valuable and employers are eager to keep them. That way AI would result in far fewer job losses.<\/p>\n Acemoglu said pro-worker AI would be far better for productivity, social cohesion and holding down income inequality. He acknowledged, however, that pro-worker AI is \u201cnot so good for the business models of the big tech companies\u201d \u2013 their models seek to maximize profits and automation.<\/p>\n To get AI companies to embrace a pro-worker approach will undoubtedly take considerable pressure from government and society. The Biden White House held discussions with labor leaders<\/a> about how to make AI less harmful to workers. The Biden administration and various Biden-era agencies adopted several pro-worker AI policies<\/a> (for instance, to limit harmful AI surveillance<\/a>), but his administration never issued recommendations or regulations to steer AI\u2019s future in a way that would minimize layoffs. Perhaps that would have happened in a second Biden term.<\/p>\n Three days after returning to the White House, Donald Trump \u2013 with tech billionaires having helped finance his campaign and inauguration \u2013 rescinded Biden\u2019s modest efforts<\/a> to make AI less harmful. Trump essentially gave the green light to AI companies to pursue whatever strategies they wanted, workers and the public be damned. On Thursday, he issued an executive order aiming to block any state laws restricting AI.<\/p>\n \u201cThe Trump administration<\/a> has really changed the trajectory of the conversation about AI,\u201d said Amanda Ballantyne, until recently the director of the AFL-CIO\u2019s Technology Institute. (The AFL-CIO is the main US labor federation.) \u201cI wouldn\u2019t say we\u2019re seeing a great trajectory right now for what AI will mean for workers. We\u2019re living through a live-action experiment as tech companies and employers try to figure out how to use this new technology, with many people concerned that it will be destabilizing and bad for workers.\u201d<\/p>\n It would be great if the path forward for AI could follow Germany and Scandinavia\u2019s model of having industry, labor and government work together to develop policies that help business and workers alike. \u201cWe\u2019re in new territory with AI technology,\u201d said Ballantyne, now a senior fellow at New America, a progressive thinktank. \u201cWe should take lessons from deindustrialization in the United States, where our anemic policy response to the negative effects of free trade left many working people behind. We can, and absolutely should, use the government to incentivize the development of pro-worker AI and smart regulations, as we\u2019ve done previous times in history.\u201d Ballantyne pointed to Franklin Roosevelt\u2019s program that led to the electrification of hundreds of backwards rural communities, helping to improve living conditions and spur economic development.<\/p>\n \u201cThe Democrats,\u201d Ballantyne said, \u201cshould make this a core part of their platform, that it\u2019s possible to use government to create the kind of economy we want\u201d \u2013 in this case, to incentivize tech companies to develop AI in more pro-worker, less destructive ways.<\/p>\n Four decades ago, Wassily Leontief, an NYU economics professor who won a Nobel prize for economics, hypothesized, somewhat playfully, about a future with so much automation that just one factory worker would be left<\/a>, and that worker would turn a switch, and all the world\u2019s manufacturing would get done. Leontief worried that the factory owners would get all the benefits and workers would get pauperized.<\/p>\n