Pluralistic: The REAL AI automation threat to workers (11 Jan 2024)

Today's links

A sweatshop: women sit around a table sewing. Through the lone window, we can see a 'code waterfall' effect as seen in the credits of the Wachowskis' 'Matrix' movies. To their left stands a man in a pin-stripe suit, looking at his watch. His body language radiates impatience. His eyes have been replaced by the staring red eyes of HAL9000 from Kubrick's '2001: A Space Odyssey.'

The real AI automation threat to workers (permalink)

Long before the current wave of AI hype, we were being groomed for automation panics with misleading stories. Remember this one? "'Truck driver' is the most common job in America. Self-driving trucks are just around the corner. How can we prevent America's army of truckers from turning into a howling mob when the robots steal their jobs?"

It was absolute nonsense. First of all, "truck driver" isn't a particularly common job in America! The BLS lumps together all cargo vehicle drivers under a single classification. The category error here was thinking that every delivery van driver, furniture mover, and courier is behind the wheel of a big rig, cracking wise on a CB radio as they tear up the interstate.

But what about automation threats? It's possible that if we redesigned the interstates to give 18 wheelers their own separated lanes, and then set them to following one another, that they could traverse long distances in that way. Congratulations, you've just invented a shitty, failure-prone train.

"Shitty train AI" does not threaten the job of the vast number of people the BLS classifies as "truck drivers." For one thing, "shitty train AI" isn't going to pilot a UPS van around the streets of a busy city with other road users. Sure, a few robotaxi companies have bamboozled city governments into conscripting the city's residents into an uncontrolled murderbot experiment. These are not going well:

More than $100b has been set on fire chasing the robotaxi dream, and the result is most charitably described as a technological curiosity, requiring 1.5 high-waged remote technicians to replace each low-waged driver:

But even if we could perfect this technology, robots still wouldn't replace all those "truckers" who drive delivery vans (to say nothing of moving vans!). The hard part of driving a UPS van isn't just getting it from place to place – it's getting the parcel into the place. The robo-van would still need at least one person to get the parcel from the back of the van and into the reception desk, porch, or other delivery zone. It's not going to fire those parcels at your door with a catapult. It's also not going to deliver them by drones. Drone delivery is another one of those historical curiosities, capable of delivering a very narrow range of parcels, under even narrower circumstances:

If all UPS delivered was lightweight, non-fragile rectangular parcels ordered by people with large, unobstructed back yards, then sure. Congrats, you've just created the world's least-useful parcel delivery service!

All that said, the big rig drivers probably don't need to worry about robots stealing their jobs. It's not even clear that "shitty train" is within our technological grasp, but even if it is, there's yet another problem with the AI automation trucker jobpocalypse: "trucker" is already one of the worst jobs in America:

It's hard to overstate just how fucking terrible it is to be a trucker. Truckers are trapped in abusive debt holes by their employers – who misclassify their workforce as "contractors" in a bid to sidestep labor law. Shriven of any labor rights, truckers are forced into the most ghastly, body-destroying, family-wrcking, financially precarious existence imaginable.

You can drive a truck for years, give almost all of the money you earn back to your employer (who denies that you're their employee) to pay back the usurious loan for your truck. Then, your employer can underschedule for shifts so that you miss a loan payment, and they can repo your truck and keep the six-figure repayment you've already made to them, leaving you destitute.

They can force you to work for hours – days! – without pay while you wait for loading and dispatch. They can make you drive long past the point of safety, then, if (when) you get into a wreck, they can fine you for not taking the mandated rest breaks.

Now, these drivers aren't about to be replaced by AI – but that doesn't mean that AI won't affect their jobs. Commercial drivers are among the most heavily surveilled workers in the country. Amazon's drivers (whom Amazon misclassifies as subcontractors) have their eyeballs monitored by AI;

AIs monitor the voices of the (primarily Black, primarily female) workforce at Arise – homeworkers who field customer service calls for blue-chip companies like Carnival Cruises and Disney. They're listening for unruly children or pets in the background, and workers who fail to muffle these dependents lose the contracts they have to pay to train for:

And AI monitors the conduct of workers on temp-work apps. If a worker is dispatched to a struck workplace and refuses to cross the picket-line, the AI boss fires you and blacklists you from future jobs for refusing to robo-scab:

Writing in The Guardian, Steven Greenhouse describes the AI-enabled workplace, where precarious, often misclassified workers are monitored, judged, and fined by algorithms:

Whether it's the robot that gets you disciplined for sending an email with the word "union" in it or the robot that takes money out of your paycheck if you take a bathroom break, AI has come for the workplace with a vengeance.

Here's a supreme irony: nearly all of the beneficial applications for AI require that AI be used to help workers, not replace them, which is absolutely not how AI is used in the workplace. An AI that helps radiologists by giving them a second opinion might help them find tumors on x-rays, but that's a tool that reduces the number of scans a radiologist processes in a shift, by making them go back and reconsider the scans they've already processed:

But AI's sales pitch is not "Buy an AI tool and increase your costs while increasing your accuracy." The pitch for AI is "buy an AI and save money by firing workers." Given how bad AIs are at replacing humans, this is a bad deal all around, both for the worker who loses their job and the customer who gets the substandard product the AI makes.

There is a very limited slice of applications where an AI could make a lot of money for a company that deploys it, without costing that company anything when the AI screws up. For example, AI is a really good tool for fraud! Rather than paying people to churn out millions of variations on a phishing email, you can get an AI to do it. If the AI writes a bad phishing email, it's OK, since nearly all recipients of even good phishing emails delete them. What's more, no one will fine you or publish an op-ed demanding that your board of directors fire you if you buy an incompetent AI to commit fraud. Fraud is a high-value, low-consequence environment for using AI.

Another one of those applications is managing precarious workers who don't have labor rights. If the AI unfairly docks your worker's wages, or forces them to work until they injure themselves or others, or decides that their eyeball movements justify firing them, those workers have no recourse. That's the whole point of pretending that your employees are contractors: so you can violate labor law with impunity!

But that's not the ironic part. The ironic part is that "being a shitty boss" is the one AI application that companies are willing to increase their net spending on. No one buys an eyeball-monitoring AI so they can fire a manager. This is the one place where AI is there to augment, rather than replace, an employee.

This makes AI-based bossware subtly different from other forms of Taylorism, the "scientific management" fad of the early 20th century that saw management consultants choreographing the postures and movements of workers to satisfy the aesthetic fetishes of their employers:

The pseudoscientific cod-ergonomics of the 1900s was demeaning and even dangerous, but it wasn't automated, and if it increased worker output, this was incidental to the real purpose of making workers move like the machine-cogs their bosses reassured themselves they were:

Every AI panic is a way of deflecting attention from the real, grimy, here-and-now ways that AI is destroying our lives by demanding that we entertain nonsensical science fiction claims about large, shiny existential risks that AI might present in the future.

The "X-risk" of the spicy autocomplete chatbot waking up and using its newfound sentience to turn us all into paperclips is nonsense. Adding words to the plausible sentence generator doesn't turn it into a superintelligence for the same reason that selectively breeding faster horses doesn't lead to locomotives:

But there is a way that AI could destroy the human race! The carbon footprint and water consumption associated with training and operating large-scale models are significant contributors to the climate emergency, which threatens the habitability of the only planet in the known universe capable of sustaining human life:

Likewise, AI isn't going to replace you at work. But it's already augmenting your shitty boss's ability to rip you off, torment you, maim you and even kill you in order to eke out a few more basis points for the next shareholder report.

Science fiction is a fun and useful way to tell parables about our current technologies. But it's not a roadmap for the future. The fact that sf writers like me found AIs as useful metaphors to describe Earth's dominant artificial life form – the limited liability corporation – doesn't mean that superhuman AIs should – or can – be created.

(Image: Cryteria, CC BY 3.0, modified)

Hey look at this (permalink)

A Wayback Machine banner.

This day in history (permalink)

#15yrsago Pneumatic tube-based postal systems of the late 19th century

#15yrsago Another London photographer arrested for “terrorism” (i.e. “taking a picture of a public building”)

#10yrsago Trans-Pacific Partnership: how the US Trade Rep is hoping to gut Congress with absurd lies

#10yrsago Baybrook Remodelers sue (and sue and sue) people who give them negative reviews

#10yrsago Chronology of Canadian Tories’ war on science libraries

#5yrsago Vizio exec: we’d have to charge a premium on “dumb” TVs to make up for the money we’ll lose by not spying on you

#5yrsago Big Data’s “theory-free” analysis is a statistical malpractice

#5yrsago Court strikes down Iowa’s unconstitutional ag-gag law

#5yrsago McGingerbread Hell: bakers celebrate McMansion Hell with delicious monster houses

#5yrsago Bird Scooter tried to censor my Boing Boing post with a legal threat that’s so stupid, it’s a whole new kind of wrong

#1yrago Deena Mohamed's 'Shubiek Lubiek'

Colophon (permalink)

Today's top sources:

Currently writing:

  • A Little Brother short story about DIY insulin PLANNING

  • Picks and Shovels, a Martin Hench noir thriller about the heroic era of the PC. FORTHCOMING TOR BOOKS JAN 2025

  • The Bezzle, a Martin Hench noir thriller novel about the prison-tech industry. FORTHCOMING TOR BOOKS FEB 2024

  • Vigilant, Little Brother short story about remote invigilation. FORTHCOMING ON TOR.COM

  • Spill, a Little Brother short story about pipeline protests. FORTHCOMING ON TOR.COM

Latest podcast: The Internet’s Original Sin

Upcoming appearances:

Recent appearances:

Latest books:

Upcoming books:

  • The Bezzle: a sequel to "Red Team Blues," about prison-tech and other grifts, Tor Books, February 2024

  • Picks and Shovels: a sequel to "Red Team Blues," about the heroic era of the PC, Tor Books, February 2025

  • Unauthorized Bread: a graphic novel adapted from my novella about refugees, toasters and DRM, FirstSecond, 2025

This work – excluding any serialized fiction – is licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commercially, provided that you attribute it to me, Cory Doctorow, and include a link to

Quotations and images are not included in this license; they are included either under a limitation or exception to copyright, or on the basis of a separate license. Please exercise caution.

How to get Pluralistic:

Blog (no ads, tracking, or data-collection):

Newsletter (no ads, tracking, or data-collection):

Mastodon (no ads, tracking, or data-collection):

Medium (no ads, paywalled):

Twitter (mass-scale, unrestricted, third-party surveillance and advertising):

Tumblr (mass-scale, unrestricted, third-party surveillance and advertising):

"When life gives you SARS, you make sarsaparilla" -Joey "Accordion Guy" DeVilla