Andrew Yang Warns That AI Is Set to Displace Millions of Workers, America Is ‘Terrible at Retraining’ Workers… ‘The Coal Miners Did Not Become Coders’

"We are terrible at retraining," Andrew Yang told CNBC on Tuesday, Aug. 18. "The coal miners did not become coders. The warehouse workers did not become nurses. It's all just ludicrous talking points that politicians say and then walk away while there's a mess left."

It is a line he has been using in one form or another since his 2020 presidential run. But since then, Yang's arguments have gone from far-out theories to everyday concerns for millions of Americans. 

The proposal is a swap. Stop taxing employment, start taxing AI use, and route the proceeds to consumers as cash. "We're subsidizing a technology that's going to replace millions of American workers," he said. "I was speaking in Michigan. I said, 'Okay guys, here's the math. We're going to spend $3 trillion on AI, and it's not going to affect anyone's jobs? I mean, that's ludicrous.'"

Sponsored Content: Meeting AI’s 8,000% energy demand growth. Invest in Frontieras before the August 27 deadline.

Strip out the politics, and there is a cost-accounting observation underneath that any CFO would recognize. "Because right now, if I hire a worker, I'm going to pay a lot in payroll taxes and healthcare and the rest of it," Yang said. "Whereas if I use AI, I don't pay any of those things." A human hire carries payroll tax, benefits, and administrative overhead on top of wages. Inference costs carry none of it, and inference is getting cheaper every year while wages do not. Whatever anyone thinks of a tax, the relative-price argument is arithmetic… and it is the mechanism sitting behind margin expansion across most of the companies buying Nvidia's (NVDA) hardware.

The detail worth pausing on is that Yang says he did not come up with it. "First, it's not even original," he said. "Dario Amodei, the CEO of Anthropic, a number of months ago, raised his hand and said, ‘You should tax us; put in a token tax.’ And he's right." An executive at a frontier AI lab asking to be taxed is not the usual shape of this debate, and Yang volunteering that the idea belongs to someone else is not the usual shape of a policy pitch.

Sponsored Content: Thanks to Jeff Bezos, you can now become a landlord for as little as $100 — and no, you don't have to deal with tenants or fix freezers. Here’s how.

For scale, Yang pointed at a single occupation: "There are 2.9 million Americans who work in call centers and customer service centers right now," he said. That's one of the categories where automated systems are furthest along and where displacement, if it comes, will show up in headcount lines rather than in think pieces.

The other side of this argument has been made forcefully by people with rather more money at stake, and it deserves the same space. Jeff Bezos has taken the opposite view outright, framing AI as a tool that multiplies what a worker can do rather than a replacement for the worker. Somebody is about to “hand you a bulldozer” after you have been digging with a shovel. That has been the historical pattern with most general-purpose technology, and it is not a naive position.

It is also not a settled one inside the industry. Sam Altman has conceded that OpenAI was “pretty wrong” about AI and jobs, which is a notable revision from the person running the company that started the current cycle. The people building these systems do not agree with each other about what they do to employment, and a reader is entitled to hold both the bulldozer argument and Yang's payroll-cost argument at the same time.

What Yang is describing has some early legislative existence. The AI Tax and Work Protection Act was only just proposed this month, but a change of this kind would likely require rewriting how employment and computing are treated in the tax code simultaneously.

For now, that leaves the useful part, which is the price signal rather than the policy. As long as hiring a person carries payroll tax and benefits and running a model does not, the tax code puts a thumb on the scale in one direction. That is true regardless of whether anyone ever does anything about it, and it is the version of Yang's argument that survives being disagreed with.