Pluralistic: 09 Oct 2022 $100 billion later, autonomous vehicles are still a car-wreck

Today's links

A chart illustrating the Gartner hype-cycle; racing down the slope from the 'peak of inflated expectations' to the 'trough of disillusionment' is the staring eye of HAL 9000 from 2001: A Space Odyssey, chased by speed-lines.

$100 billion later, autonomous vehicles are still a car-wreck (permalink)

Autonomous vehicles were always a shell-game. The last time I wrote about them was a year ago, when Uber declared massive losses. Uber's profitability story was always, "Sure, we're losing money now, but once we create self-driving cars, we can fire our drivers and make a bundle."

But Uber never came close to building an AV. After blowing $2.5b, the company invented a car whose mean-distance-to-fatal-crash was half a mile. Uber had to pay another company – $400 million! – to take the self-driving unit off its hands.

It's tempting to say that Uber just deluded itself into thinking that AVs were a viable, near-term technology. But $2.5b was a bargain, because it allowed the company's original investors (notably the Saudi royals) to offload their Uber shares on credulous suckers when the company IPOed.

Likewise Tesla, a company that has promised fully self-driving autonomous vehicles "within two years" for more than a decade. The story that Teslas will someday drive themselves is key to attracting retail investors to the company.

Tesla's overvaluation isn't solely a product of the cult of personality around Musk, nor is it just that its investors can't read a balance-sheet and so miss the fact that the company is reliant upon selling the carbon-credits that allow gas-guzzling SUVs to fill America's streets.

Key to Tesla's claims to eventual profitability was that AVs would overcome geometry itself, and end the Red Queen's Race whereby adding more cars to the road means you need more roads, which means everything gets farther apart, which means you need more cars – lather, rinse, repeat.

Geometry hates cars, but Elon Musk hates public transit (he says you might end up seated next to "a serial killer"). So Musk spun this story where tightly orchestrated AVs would best geometry and create big cities served speedy, individualized private vehicles. You could even make passive income from your Tesla, turning it over to drive strangers (including, presumably, serial killers?) around as a taxicab.

But Teslas are no closer to full self-driving than Ubers. In fact, no one has come close to making an AV. In a characteristically brilliant and scorching article for Bloomberg, Max Chafkin takes stock of the failed AV project:

Chafkin calculates that the global R&D budget for AVs has now exceeded $100 billion, and demonstrates that we have next to nothing to show for it, and that whatever you think you know about AV success is just spin, hype and bullshit.

Take the much-vaunted terribleness of human drivers, which the AV industry likes to tout. It's true that the other dumdums on the road cutting you off and changing lanes without their turn-signals are pretty bad drivers, but actual, professional drivers are amazing. The average school-bus driver clocks up 500 million miles without a fatal crash (but of course, bus drivers are part of the public transit system).

Even dopes like you and me are better than you may think – while cars do kill the shit out of Americans, it's because Americans drive so goddamned much. US traffic deaths are a mere one per 100 million miles driven, and most of those deaths are due to recklessness, not inability. Drunks, speeders, texters and sleepy drivers cause traffic fatalities – they may be skilled drivers, but they are also reckless.

But even the most reckless driver is safer than a driverless car, which "lasts a few seconds before crapping out." The best robot drivers are Waymos, which mostly operate in the sunbelt, "because they still can’t handle weather patterns trickier than Partly Cloudy."

Waymo claims to have driven 20m miles – that is, 4% of the distance we'd expect a human school-bus driver to go before having a fatal wreck. Tesla, meanwhile, has stopped even reporting how many miles its autopilot has mananged on public roads. The last time it disclosed, in 2019, the total was zero.

Using "deep learning" to solve the problems of self-driving cars is a dead-end. As NYU psych prof Gary Marcus told Chafkin, "deep learning is something similar to memorization…It only works if the situations are sufficiently akin."

Which is why self-driving cars are so useless when they come up against something unexpected – human drivers weaving through traffic, cyclists, an eagle, a drone, a low-flying plane, a deer, even some pigeons on the road.

Self-driving car hucksters call this "the pogo-stick problem" – as in "you never can tell when someone will try to cross the road on a pogo-stick." They propose coming up with strict rules for humans to make life easier for robots.

But as stupid as this is, it's even stupider than it appears at first blush. It's not that AVs are confused by pogo sticks – they're confused by shadows and clouds and squirrels. They're confused by left turns that are a little different than the last left turn they tried.

If you've been thinking that AVs were right around the corner, don't feel too foolish. The AV companies have certainly acted like they believed their own bullshit. Chafkin reminds us of the high-stakes litigation when AV engineer Anthony Levandowski left Google for Uber and was sued for stealing trade secrets.

The result was millions in fines (Levandowski declared bankruptcy) and even a prison sentence for Levandowski (Trump pardoned him, seemingly at the behest of Peter Thiel and other Trumpist tech cronies). Why would companies go to all that trouble if they weren't serious about their own claims?

It's possible that they are, but that doesn't mean we have to take those claims at face-value ourselves. Companies often get high on their own supplies. The litigation over Levandowski can be thought of as a species of criti-hype, Lee Vinsel's extraordinarily useful term for criticism that serves to bolster the claims of its target:

Another example of criti-hype: the claims about the risks of ubiquitous drone delivery – which, like AVs, are half-bullshit, half self-delusion:

Today, Levandowski has scaled back his plans to build autonomous vehicles. Instead, he's built autonomous dump-trucks that never leave a literal sandbox, and trundle back and forth on the same road all day, moving rocks from a pit to a crusher.

$100 billion later, that's what the AV market has produced.

(Image: Cryteria, CC BY, modified; Gartner, modified)

Hey look at this (permalink)

This day in history (permalink)

#20yrsago On the road with the Internet Bookmobile

#20yrsago Eldred hearing liveblog

#20yrsago GI Joe’s dream house

#20yrsago Matt Haughey on Eldred and the Commons

#15ysago Baby-naming, XKCD style

#15yrsago All the Beatles’ UK albums sped up 800% into a 1 hour MP3

#15yrsago Yahoo Music to record execs: No more DRM, ever

#15yrsago Hauntrod Funnycar for your short-range travel needs

#10yrsago Zee Avi, skewered on an angelic voice

#10yrsago Backyard Blockbusters: documentary about fan films

#10yrsago Fact-checking the RIAA’s claim that the number of working musicians fell by 41%

#10yrsago Microsoft claims ownership of the number 45, asks Google to censor the US government and Bing

#5yrsago The copay assumes the 99% treat healthcare like spiteful buffet gorgers

#5yrsago London’s amazing underground infrastructure revealed in vintage cutaway maps

#5yrsago Antifascist Greek protesters say they were tortured by police

#5yrsago Pinkwater’s Bushman Lives: absurdist misfit story is an insightful treatise on art

#5yrsago Distinguished scientist on the mistakes pundits make when they predict the future of AI

#5yrsago Triangulation is dead: what does “socialism” mean in the 21st century?

#1yrago Facebook shouldn't be in charge of how you use Facebook: Unfollow Everything and the need for (good) tech regulation

Colophon (permalink)

Today's top sources: Naked Capitalism (

Currently writing:

  • The Bezzle, a Martin Hench noir thriller novel about the prison-tech industry. Friday's progress: 508 words (47676 words total)

  • The Internet Con: How to Seize the Means of Computation, a nonfiction book about interoperability for Verso. Friday's progress: 531 words (44110 words total)

  • Picks and Shovels, a Martin Hench noir thriller about the heroic era of the PC. (92849 words total) – ON PAUSE

  • A Little Brother short story about DIY insulin PLANNING

  • Vigilant, Little Brother short story about remote invigilation. FIRST DRAFT COMPLETE, WAITING FOR EXPERT REVIEW

  • Moral Hazard, a short story for MIT Tech Review's 12 Tomorrows. FIRST DRAFT COMPLETE, ACCEPTED FOR PUBLICATION

  • Spill, a Little Brother short story about pipeline protests. FINAL DRAFT COMPLETE

  • A post-GND utopian novel, "The Lost Cause." FINISHED

  • A cyberpunk noir thriller novel, "Red Team Blues." FINISHED

Currently reading: Analogia by George Dyson.

Latest podcast: Sound Money

Upcoming appearances:

Recent appearances:

Latest books:

Upcoming books:

  • Red Team Blues: "A grabby, compulsive thriller that will leave you knowing more about how the world works than you did before." Tor Books, April 2023

This work 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):

(Latest Medium column: "Bankruptcy protects fake people, brutalizes real ones"

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