Pluralistic: 04 Jun 2020

Today's links

Minneapolis racists threaten arson over BLM signs (permalink)

An anonymous person (or persons) has been terrorizing people in Minneapolis with notes threatening to murder them through arson in retaliation for lawn signs supporting Black Lives Matter.

The notes say things like "Your neighbors are sick of riots and your SJW shit," and “Your sign 'bullshit matters' comes down or you and your home will burn real quiet while you sleep in it!"

The Roseville Police Department is investigating the threats; however, the Ramsey County Sheriff's Department has not yet been formally notified about similar notes in Falcon Heights.

White House recants "antifa bricks" conspiracy (permalink)

If you are follow the uprisings across America, you have probably encountered both right- and left-wing commentators speculating that piles of bricks at the sites of the protests were "staged" – either by "antifa" or by right-wing accelerationists or agents provocateurs.

The White House released a supercut of bricks at protest sites, claiming evidence that "antifa" provocateurs were hoping to create vandalism and violence, but they were forced to retract it after it was debunked (but not before it had 1m+ views).

The White House removed the video, but The Intercept saved a copy. You can watch it here:

And though the incidents in the video were thoroughly debunked, it does raise the question, "Why are there so many piles of bricks at these protest sites?"

The answer's pretty mundane: cities have a lot of routine construction, which requires piles of bricks. Protests are happening all over cities. Some of those protests will be near piles of bricks (it would be weird if no protest ever happened near piles of bricks).

You could produce a supercut of protesters near any urban commonplace: Starbucks, park benches, tall buildings, liquor stores, churches, McDonald's restaurants… Anything that is widespread in a city will overlap with protests in that city.

What's more, it's not outlandish to assume that cops or far-right extremists (including/especially the "Boogaloo" accelerationists) are infiltrating protests and hoping to incite violence.

American law enforcement has used agents provocateurs against protests for as long as there have been protests:

And white nationalists have been caught posing as Black Lives Matter activists on Twitter in order to incite violence.

Infiltration is a devastating tactic, because it derails every conversation into a "Are you a bot/provocateur/cop?" arguments and sows mistrust among people whose only hope of effecting change is their social cohesion.

Putin's brilliant and terrible media strategist Vladislav Surkov (purged from the administration earlier this year) made a point of boasting that he funded some of the opposition groups, without ever saying which.

The brilliant (terrible) thing about this tactic is that it works just as well if he lies as if he's telling the truth – any time two opposition groups would contemplate acting in concert, each would be unable to shake their suspicion that the other was patsies or worse.

And anyone who thinks about joining an opposition group will always have a lingering doubt about the group's authenticity, and whether they are being duped into serving as a useful idiot.

Within groups, vulnerable people who act in unusual ways – because they have anxiety or depression, or because they are neuroatypical – get isolated because their behavior is interpreted as evidence that they are plants.

So it's not weird to imagine that there are inauthentic members of your group, because it's not weird to imagine that powerful people would use such an effective, low-cost countermeasure against you.

Nor is it hard to understand why the White House would promulgate unfounded theories about provocateurs – even if they have to walk them back later. Consciously or unconsciously, Trump and his people know that instilling doubt serves their purpose for a long time to come.

Rave for "Poesy the Monster Slayer" (permalink)

No matter how many books I write (20+ now!), the first review for a new one is always scary. That goes double when the book is a first as well – like Poesy the Monster Slayer, my first-ever picture book, which comes out from First Second on Jul 14.

So it was with delight and relief that I read Publishers Weekly's (rave) review of Poesy:

"Some children fear monsters at bedtime, but Poesy welcomes them. Her pink 'monster lair' features gothic art and stuffed animals, and she makes her father read The Book of Monsters from cover to cover before lights out. 'PLEASE stay in bed tonight,' he pleads as he leaves, but there's no chance: the werewolf who soon enters her window is the size of a grizzly. 'Werewolves HATED silver,' Poesy knows, 'and they feared the light' 0 armed with a Princess Frillypants silver tiara and a light-up wand, she vanquishes the beast. And that's just the beginning of her tear through monsterdom. 'Poesy Emmeline Russell Schnegg,' her mother growls from the doorway (in a funny turn, the girl gains a middle name every time a parent appears). Assured panels by Rockefeller (Pop!) combine frilly with threatening, illuminated by eerie light sources. Doctorow, making his picture book debut, strikes a gently edgy tone ('He was so tired,' Poesy sees, 'that he stepped on a Harry the Hare block and said some swears. Poor Daddy!'), and his blow-by-blow account races to its closing spread: of two tired parents who resemble yet another monster. Ages 4-6."


I had planned to do a launch party at Dark Delicacies, my neighborhood horror bookstore, on Jul 11, but that's off (obviously).

So we're doing the next-best thing: preorder from the store and you'll get a signature and dedication from me AND my daughter, Poesy (the book's namesake).

Autodetecting autogenerated text (permalink)

GPT2 is machine learning "language model" from Openai: feed it some articles and it'll spit out a convincing new, very plausible article, untouched by human hands.

Its release was attended by enormous hype, even by machine learning standards, largely because Openai said they were nerfing the public version to keep it from being used to create floods of undetectable fake articles that would destroy our discourse.

But the undetectability of ML forgeries is always overstated. Indeed, the same tools used to create these forgeries are also very good at detecting them.

Enter "The Giant Language Model Test Room."

GLTR provides an example of why sometimes, "The answer to the machine is the machines." As its authors showed in 2019 when they published a video, conference paper, and sourcecode, you can use the same statistical techniques as GPT2 to detect texts generated by GPT2.

GPT2 works a little like autocomplete: after each word, it tries to guess what the most likely next word would be based on its training data.

So texts created by GPT2 are "unsurprising" in a mathematical sense: the words that follow one another are not odd juxtapositions, they're statistically likely successors.

It's an example of the fundamental conservativeness of machine learning:

GLTR performs a statistical analysis of each word in a text and colors it based on how common it is in the context of the word that came before it: green/yellow for predictable; purple/red for "surprising."

You can look at the colored output and tell, at a glance, how "normal" and thus machinelike, it is (alas, this doesn't work for people with colorblindness).

It's conceivable that you could counter this by programming GPT2 to pick "weird" words at random intervals. But because GPT2 doesn't understand language – it merely performs statistical analysis – it can't know whether these weird words will make its sentences incomprehensible.

Though a human being can, which raises the possibility of using GPT2 as a "centaur" (a human-computer collaboration team): a human could tweak computer-generated forgeries to make them both comprehensible and statistically undetectable.

Or vice-versa: if you wanted to remove traces of distinctive style to combat "stylometry" (identifying anonymous authors with stats, by analyzing texts for stylistic quirks), you could use this to remove the "surprising" elements from your work.

AMC's bankruptcy is the first domino in monopolist collapse (permalink)

AMC is (still) going bankrupt.

The company is blaming covid for its demise, but as with so many pandemic deaths, the cause of death is complicated by comorbidities that left it in a weakened and vulnerable state.

AMC was repeatedly looted by private equity firms that loaded it up with debt, borrowing money they paid to themselves as fees and dividends. First, they were bled by China's Wanda Group, then the notorious looters of America's Silverlake Partners.

When AMC says they can't pay their bills because they've had to close their doors, a substantial part of those bills are the company's debts, which were not incurred to invest in improving theaters or buying their buildings (which would shield them from rent shocks).

No, that money was funnelled to billionaires who style themselves "job creators."

What's more, AMC took advantage of America's nonexistent antitrust law to buy most of its competitors, making it the only screen in town for many of America's cities.

AMC's demise matters, then, because it constitutes the bulk of American screens.

The company's leading suitor is another monopolist, Amazon, flush with cash thanks to pandemic profiteering and shaving costs by endangering its employees' lives:

And that matters to another important monopolist, Disney, which, like Amazon and AMC, has bought many of its competitors, notably Fox, which it acquired in 2019.

To a first approximation, Disney now IS the American cinematic industry. 80% of last year's box office hits were owned by Disney.

If you were wondering why Universal didn't care that AMC announced that it would no longer screen Universal movies in retaliation for Universal's offering home streaming of first-run movies, that's why. Basically, no one but Disney is in the cinematic release business.

So Disney needs big screens, especially since its Plan B, streaming media through Disney+, is being demolished by Time-Warner/AT&T;, which is using Trump FCC Chairman Ajit Pai's obliteration of Net Neutrality to upcharge customers who choose Disney+ over HBO Max.

And of course, Universal owns Comcast, so it's poised to do the same.

Some people watched the endless mergers and industry concentration of the past 40 years and thought, "Well, at least I'll get a titan who'll protect me and give me the things I love."

But that's not how concentrated power works – instead, it proceeds with brutal brinksmanship that destroys the things we love, the jobs those things created, and the businesses behind those jobs.

This day in history (permalink)

#15yrsago I have seen God in a cup of chocolate

#10yrsago Open garage-doors in less than a minute with a hacked kid's toy

#5yrsago There was a David Cameron who swallowed a fly…

#1yrago Magic for Liars: Sarah Gailey's debut is a brilliant whodunnit in the vein of The Magicians

#1yrago Why is there so much antitrust energy for Big Tech but not for Big Telco?

#1yrago Empirical analysis of behavioral advertising finds that surveillance makes ads only 4% more profitable for media companies

#1yrago European legal official OKs orders that force Facebook to globally remove insults to politicians like "oaf" and "fascist" (as well as synonyms)

#1yrago Blood testing giant Quest Diagnostics lost 12,000,000 patients' personal, financial and medical data

#1yrago The New York Privacy Act goes even farther than California's privacy legislation

#1yrago Joe Biden repeatedly claimed to have marched for civil rights. He didn't.

Colophon (permalink)

Today's top sources: Four Short Links (, Slashdot (, Naked Capitalism (

Currently writing: My next novel, "The Lost Cause," a post-GND novel about truth and reconciliation. Yesterday's progress: 588 words (23172 total).

Currently reading: Adventures of a Dwergish Girl, Daniel Pinkwater

Latest podcast: How Big Tech Monopolies Distort Our Public Discourse

Upcoming appearances:

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

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

"Little Brother/Homeland": A reissue omnibus edition with a new introduction by Edward Snowden:

This work licensed under a Creative Commons Attribution 4.0 license. That means you can use it any way you like, including commerically, 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):

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

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.