Senate report says AI will take 97M US jobs in the next 10 years, but those numbers come from ChatGPT

Bernie Sanders calls for a robot tax and a 32-hour work week in response

ai-pocalypse A US Senate committee led by Bernie Sanders (I-VT) has produced a report claiming that about 97 million US jobs could be lost to AI and automation over the next decade. There's just one problem: it got those figures from ChatGPT.

According to the The US Senate Committee on Health, Education, Labor, and Pensions (HELP) report [PDF], artificial intelligence and automation could displace tens of millions of American jobs over the next decade. Staff compiled the report by reviewing economic and corporate data...then asking ChatGPT to analyze federal job descriptions and estimate which occupations are most vulnerable to replacement.

As a result, the data looks hallucinogenic. Here are the figures showing how many humans AI is expected to make redundant:

• 89 percent of fast food and counter workers
• 83 percent of customer service representatives
• 81 percent of laborers and freight, stock, and material movers
• 80 percent of secretaries and administrative assistants
• 76 percent of bookkeeping, accounting, and auditing clerks
• 76 percent of stockers and order fillers
• 66 percent of office clerks
• 65 percent of teaching assistants
• 64 percent of auditors and accountants

"The same handful of oligarchs who have rigged our economy for decades — Elon Musk, Larry Ellison, Mark Zuckerberg, Jeff Bezos and others — are now moving as fast as they can to replace human workers with what they call ‘artificial labor,’" said Sanders, who is the ranking member of HELP. "If we do not act, the result could be economic devastation for working people across this country."

As an answer to this perceived jobs apocalypse, he proposes that the government should institute a "robot tax" to be paid when workers are replaced, a 32-hour working week for everyone else without reducing pay, a requirement for corporations to share profits with workers and allow them on the board of directors, a ban on union busting, and another ban on stock buybacks by corporations.

But these figures are, frankly, nonsensical. AI can do many things almost as well as humans, but food service isn't one of them, nor is manual labor, and if you're trusting AI to handle your accounts, the SEC would probably like a word. On the other hand, Sanders warned that robotics could be taking over from humans.

"Elon Musk has said he wants Tesla to build millions of robots. And what will these robots do? They will replace the men and women working in factories, warehouses and in restaurants. That means millions of good jobs could disappear. It’s not complicated," Sanders said in an op-ed on Fox News, not the usual place you'd find the democratic socialist.

But despite Elon Musk's claims about robots being able to do the jobs of humans, there's little evidence yet that this is possible in the short term. Yes, Amazon is using robots for heavy lifting in its warehouses, for example, but they fall well short of human staff.

A recent Yale study also suggested that there's no evidence of a "discernible" disruption in the labor market due to AI. However, some companies, such as Salesforce and Fiverr, have cited AI as the cause of recent layoffs.

The report claims that companies are laying off staff, replacing them with AI systems, and reaping huge profits (although the jury is still out on that). While there's some real danger down the road, the suggested remedies seem to bear little relation to the real-world situation that many workers find themselves in.

Ironically, some of the recommendations in the report make good sense, even if they have nothing to do with AI or robotics. If the government banned stock buybacks — which were illegal until 1982 — it could encourage firms to invest in production and innovation instead of rewarding shareholders and supporting the share price. Intel, Boeing, and others spent billions on such schemes (which also boost CEO pay) and then went to the government for bailouts when times got hard.

But fearmongering over AI with dodgy data isn't going to fix that problem, nor the chronic underinvestment in US manufacturing. Instead this looks like the standard talking points that Sanders has been banging on about for years. Increasing union representation, for example, wouldn't be a bad idea in some trades, but how HELP thinks this would solve AI issues isn't remotely clear.

And, while adding paid family leave, as Sanders also suggests, would be helpful to workers, it does nothing to prevent them from being automated out of a job. If we are facing an AI-pocalypse, then specific solutions are needed, not just the usual grab-bag of old policies. ®