This post was originally published on this site.
Over the weekend, a knee‑high delivery rover rolled past my neighborhood coffee shop in Los Angeles, its bright white “eyes” blinking as it threaded between a mom with a stroller and a dog. Neither the mom, the baby, or the dog paid any attention to the robot. It was just normal. And yet, it was not so long ago that people burst out of cafés, phones aloft, to film these rolling delivery robots, taking selfies with them, or just gawking as a real live robot entered the universe. Today I was the only person looking up from my cappuccino. William Gibson’s line that “the future is already here, it’s just not evenly distributed” has never felt so literal. In other words, the job loss to robots isn’t just coming, it’s already started.
Consider the humble intern. A survey conducted by Intelligent.com found that 70% of hiring managers believe artificial intelligence can do an intern’s job—and that was last year. Back then, in April of 2024, 5% of respondents said they once offered internships but no longer do; nearly a third of that subset blamed AI and its ability to handle the busywork once assigned to 22‑year‑olds who were just happy to be there. A year later, the numbers are staggering: The Indeed Hiring Lab’s latest Labor Market Update reported earlier this month that internship postings as of April 3 slipped below their 2019 baseline and are now 11 percentage points lower than the same date in 2024, a stark signal that on‑ramp roles are evaporating before the class of 2025 has even collected a diploma.
People in Silicon Valley have called this current job replacement “head‑count arbitrage,” which is a tech-speak way of saying, “we don’t need people anymore!” Research assistants? Language models can now ingest 200‑page market reports, cross‑check citations, draft a PowerPoint, and pull off entire research projects—all done in the cloud and delivered to your computer in a few minutes at most. Those customer service calls where you’re screaming at the phone and yelling “I said operator!!!” are also likely going to be a thing of the past. Ikea is already replacing such support roles with a new AI bot called Billie that can help you with missing parts for your bookshelf or scheduling deliveries. Copywriting, storyboard design, personal‑finance advice, and a long list of other jobs are either already being replaced by AI or on the verge of being replaced.
The same dynamic is spreading into all areas of the labor pool. In the Permian Basin, Chevron is using AI technology at drilling rigs that detect early signs of pressure changes faster than humans can, and the company boasts, “AI helps Chevron extract more oil for less.” JPMorgan Chase has deployed AI to analyze commercial-loan agreements in seconds—work that used to take lawyers and loan officers 360,000 hours a year to complete. More than 40 years’ worth of work now done in seconds. Radiology departments are implementing AI systems that can spot anomalies in medical images with accuracy rivaling human specialists. Legal research, once performed by armies of paralegals and junior associates, is increasingly automated. Even creative fields are being replaced by AI already—advertising agencies now use AI to help with generating campaign concepts, copywriting, and visual designs that previously required teams of human creatives, including Coca-Cola, Heinz, and MintMobile.
Yet the most tangible proof that the human workforce is on notice rides on wheels, and, soon enough, legs. Los Angeles joined San Francisco this spring as the newest playground for Waymo’s driverless taxis. The initial thrill of sliding into a car with no driver evaporates fast. Everyone I know who has taken a Waymo says the same thing: they will never take a human-driven Uber again. “I don’t have to talk to anyone,” a friend crowed after her maiden voyage. “I can fart, pick my nose, do whatever I want, and the driver never gets offended. And honestly, it feels safer than a human who’s checking texts.” She added: “And you don’t have to tip!”
Two years ago, Goldman Sachs estimated that generative AI alone could automate the equivalent of 300 million full‑time jobs worldwide. Goldman also predicted that AI replacing humans could raise the global GDP by 7%, so I guess there is an upside, depending on who that 7% is distributed to. McKinsey predicted in a 2017 report that potentially hundreds of millions of workers would lose their jobs to automation by 2030. Pick any timeline you like, the trajectory is clear: first tasks, then jobs, then entire career paths are all going to be replaced by machines, and it’s already happening.
Optimists in tech keep trying to insist that each technological leap eventually creates more positions than it destroys: the cotton gin begot the textile engineer, the ATM spawned armies of financial‑product designers. That may be true in the long sweep of history, but it’s not true today. Jobs are vanishing en masse, and with the limited exception of prompt inputter, any new job that is created could also be replaced by AI before anyone has a chance to master their new field.
The next 12 months are going to be astounding when it comes to the speed at which AI will get smarter, faster, and more capable of doing what we do, but better. OpenAI, Google, Anthropic, and Meta are releasing larger, sharper models at a cadence to rival Donald J. Trump’s new lists of tariffs, while well‑funded Chinese companies, some that are allegedly even clones of US-based AI companies, are quickly catching up to American innovation. Chipmakers can’t fabricate silicon fast enough. Venture capitalists, giddy at the prospect of permanent cost savings, are flinging billions at any start‑up whose deck pairs “GPT‑4o” and “head‑count arbitrage” on slide three.
Which brings me—gulp!—to my own line of work. If a language model can synthesize survey data, sprinkle in a Gibson quote, and spit out a wry thousand‑word column in flawless AP style, what exactly am I contributing besides carpal tunnel syndrome and an affection for em dashes? Vanity Fair may soon discover that its tech reporter can be replaced by an algorithm that never complains about edits, never files an expense report, and doesn’t call in sick—ever. Because if the intern is really the canary in the coal mine, I’m not far behind.