Pluralist, a daily link-dose: 20 Feb 2020

Today's links

  1. The 2020 Nebula Award Finalists: a bumper crop of outstanding SF
  2. Uber driver/sharecroppers drive like maniacs to make quota: subprime lending + gig economy = stay off the roads
  3. Barclay's bankers forced to endure nagging work-computer spyware: the shitty technology adoption curve at work
  4. Bernie Sanders leads in 10 out of 10 polls: but unless he can get a majority of pledged delegates, he'll be ratfucked by superdelegates
  5. Bloomberg: kids only like Sanders because they're stupid: "Because our kids no longer learn civics in school they longer study Western history, they no longer read Western literature…"
  6. "Secure erase" with a bolt-cutter: Jamie Zawinski doesn't mess around when it comes to getting rid of old hard drives.
  7. Adding 2 inches of tape to a road-sign induces sudden 50mph acceleration in Teslas: Adversarial examples are unstoppable.
  8. Colophon: Recent publications, current writing projects, upcoming appearances, current reading

The 2020 Nebula Award Finalists

Ooh, they've announced the Nebula Award finalists! It's a pretty fucking GREAT roster! Congrats to all the nominees! Go you!

Best novel:

  • Marque of Caine, Charles E. Gannon
  • Ten Thousand Doors of January, Alix E. Harrow
  • A Memory Called Empire, Arkady Martine
  • Gods of Jade and Shadow, Silvia Moreno-Garcia
  • Gideon the Ninth, Tamsyn Muir
  • A Song for a New Day, Sarah Pinsker

Best novella:

  • Anxiety Is the Dizziness of Freedom, Chiang
  • The Haunting of Tram Car 015, Djèlí Clark
  • This Is How You Lose the Time War, Amal El-Mohtar & Gladstone
  • Her Silhouette, Drawn in Water, Vylar Kaftan
  • The Deep, Rivers Solomon et al
  • Catfish Lullaby, AC Wise

Best novelette:

  • A Strange Uncertain Light, GV Anderson
  • For He Can Creep, Siobhan Carroll
  • His Footsteps, Through Darkness & Light, Mimi Mondal
  • The Blur in the Corner of Your Eye, Sarah Pinsker
  • Carpe Glitter, Cat Rambo
  • The Archronology of Love, Caroline M Yoachim

Best short story:

  • Give the Family My Love, AT Greenblatt
  • The Dead, In Their Uncontrollable Power, Karen Osborne
  • And Now His Lordship Is Laughing”, Shiv Ramdas
  • 10 Excerpts from an Annotated Bibliography on the Cannibal Women of Ratnabar Island, Nibedita Sen
  • A Catalog of Storms, Fran Wilde
  • How the Trick Is Done”, AC Wise

Full roster and details on the Nebula Awards Weekend (Los Angeles, May 28-31) here:

Uber driver/sharecroppers drive like maniacs to make quota

When Imran Khan got into an Uber, his driver explained that the reason all the other Ubers in their traffic jam were driving so unsafely is that they're sharecroppers hiring their cars from a millionaire who won't pay them until they make quota.

The drivers are too economically precarious to lease cars on their own, so this guy acts as a subprime lender, and part of his deal is that the payments Uber sends to the drivers actually get diverted to his bank account, and they don't see a penny until they hit quota.

They're people who are working fulltime jobs and then driving Uber before and after those jobs to make ends meet. So they have to (literally) cut corners if they're going to make this work, but if they get a ticket or lose points due to passenger reviews, they lose the car.

This was in DC, and the subprime loan-shark was based in Virginia, but you can imagine that it happens everywhere (Kahn's replies from passengers who've heard the same tale elsewhere suggests that this is true).

If you get run down by one of these guys, it's the market at work: their access to capital is limited by their economic situation; their wages are determined by supply and demand, they need to eat, clothe themselves, and have shelter. This is a totally predictable outcome.

Barclay's bankers forced to endure nagging work-computer spyware

Bankers at Barclays are furious that their computers have been fitted with employer-provided spyware that monitors every keystroke and nags them if they're not working hard enough. This is a great example of the Shitty Technology Adoption Curve: first we subject powerless people to bad technology and use them to normalize it even as we sand the rough edges off using their lives as sandpaper.

Then we work our way up the privilege gradient, to people with more and more social power. First it's kids, or homeless people, or refugees, or benefits recipients or criminals. Then it's blue collar workers, university students and library patrons. Finally it's low-level white-collar workers, then their managers, then, eventually CEOs (as I've often said, CCTVs that watch you while you eat went from supermax prisons to Google Home in a generation).

"The system tells staff to “avoid breaks” as it monitors their productivity in real-time, and records activities such as toilet visits as 'unaccounted activity'."

"It added: 'Tips: mute the phone, disable email/chat pop-ups, avoid breaks for 20+ minutes, 2–3 times a day.'"

Bernie Sanders leads in 10 out of 10 polls

Ten out of ten national polls put Bernie Sanders in a commanding lead over other candidates for the Democratic leadership. However, he is unlikely to attain a majority of delegates at the DNC, meaning the "superdelegates" will get to throw out the party members' primary votes and impose an establishment candidate on the country.

Bloomberg: kids only like Sanders because they're stupid

Michael Bloomberg, 2016, Oxford University: "Young people listened to Sanders…Because our kids no longer learn civics in school they longer study Western history, they no longer read Western literature…" "We are trying to change and dumb down the system and if you don't know what happened in the past you're going to have to relive it."

Or, as Vice put it, "Bloomberg has a surprising theory about why young people love Sanders: They’re morons."

Bloomberg: "The solution to our problems is to improve education, not to try to penalize people because they are successful. If you don't have successful people you're never going to have the wherewithal to support to help those who are not. We've tried socialism, it doesn't work.”

Basically: eugenics. Some people are Atlases and we'd better not piss them off or they might shrug and leave the rest of us in the cold. Bloomberg has $64b in assets and the median US worker has $69k in assets because Bloomberg is worth 927,536 times more than that worker. Inequality is always comorbid with eugenics. If you can't admit that no one can "earn" a billion dollars, then you have to stipulate that some people are just worth a lot more than the rest of us.

"Secure erase" with a bolt-cutter

When Jamie Zawinski wants to securely erase his data, he doesn't mess about. His break-my-drive-in-half-with-a-bolt-cutter method is a lot less messy than my tried-and-true hit-it-with-a-hammer method.

Adding 2 inches of tape to a road-sign induces sudden 50mph acceleration in Teslas

McAfee security researchers stuck a 2" strip of black tape on 35mph speed limit sign so that it kinda-sorta looked like an 85mph sign, then ran autopiloting Teslas past it: they automagically accelerated by 50mph after detecting it.

McAfee reported it to Tesla and Mobileeye, who do some of the autopilot stuff, and neither vendor plans to address it.

The ML term for this is "adversarial example" – that's when you make small changes, including human-imperceptible ones, that cause otherwise reliable ML classifiers to misfire terribly.

I once had a dinner conversation with the CSO of one of the largest ML companies in the world. They confided that they believed you could never eliminate adversarial examples from classifiers, meaning they would always be vulnerable to this kind of attack. If that's right, the implications are staggering. It basically means you shouldn't use ML in any situation where someone is incentivized to trick it.

So maybe you can use it on a conveyor belt in a recycling plant to sort plastics from paper and direct a robot-arm. But almost every application for ML eventually becomes adversarial.

ML is supposedly pretty good at distinguishing precancerous moles from benign ones, which sounds non-adversarial. But consider the doctor who wants to gin up billings for unnecessary surgeries, or the insurer that wants pretences to deny necessary ones.

It doesn't take a lot of imagination to see how the trajectory of most (if not all) ML classifiers is to be in adversarial situations. If we can't provably demonstrate that a classifier is immune to adversarial examples (including ones as trivial as "2 inches of tape on a sign"), there's not a whole lot of applications for them in the long-term.


Today's top sources: Slashdot ( and Naked Capitalism (

Hugo nominators! My story "Unauthorized Bread" is eligible in the Novella category and you can read it free on Ars Technica:

Upcoming appearances:

Currently writing: I just finished a short story, "The Canadian Miracle," for MIT Tech Review. It's a story set in the world of my next novel, "The Lost Cause," a post-GND novel about truth and reconciliation.

Currently reading: I finished Andrea Bernstein's "American Oligarchs" yesterday; it's a magnificent history of the Kushner and Trump families, showing how they cheated, stole and lied their way into power. I'm getting really into Anna Weiner's memoir about tech, "Uncanny Valley." I just loaded Matt Stoller's "Goliath" onto my underwater MP3 player and I'm listening to it as I swim laps.

Latest podcast: Persuasion, Adaptation, and the Arms Race for Your Attention:

Upcoming books: "Poesy the Monster Slayer" (Jul 2020), a picture book about monsters, bedtime, gender, and kicking ass. Pre-order here:

(we're having a launch for it in Burbank on July 11 at Dark Delicacies and you can get me AND Poesy to sign it and Dark Del will ship it to the monster kids in your life in time for the release date).

"Attack Surface": The third Little Brother book, Oct 20, 2020.

"Little Brother/Homeland": A reissue omnibus edition with a very special, s00per s33kr1t intro.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.