[Plura-list] In Search Of A Flat Earth; A cryptographic mystery solved; Youtube's war on algorithmic radicalization

Cory Doctorow doctorow at craphound.com
Fri Sep 18 14:28:06 EDT 2020

Today's links

* In Search Of A Flat Earth: From Flat Earth to Qanon, the rise of
murderous cults.

* A cryptographic mystery solved: How a Cuban numbers station's
decade-long opsec fail blew Russian spies' cover.

* Youtube's war on algorithmic radicalization: Selective rabbit-hole
elimination is HARD.

* The Hardware Lottery: Chip architecture is the secret legislator of
machine learning.

* This day in history: 2015

* Colophon: Recent publications, upcoming appearances, current writing
projects, current reading


👤 In Search Of A Flat Earth

"In Search Of A Flat Earth" is Dan Olson's 1:16m documentary about the
true nature of conspiratorial thinking, AKA "Why do some people think
the Earth is flat?" (and also that Qanon is real, etc).


This is a subject I'm really interested in. I think we grossly
overemphasize the role that algorithms play, and largely ignore the role
that corruption and other real-world "conspiracies" play in making other
conspiracy theories seem plausible.


But I don't think that algorithms are irrelevant, nor do I think that
everything can be explained by the trauma of living through real-world
conspiracies - like the ones that led to wage stagnation,
deindustrialization, and skyrocketing health and housing costs.

And this is where Olson comes in: he does a fantastic and illuminating
(and often entertaining) job of showing how conspiracism is
multifactorial, and the role that fascism plays in conspiratorial
thinking and its drive to reduce complex problems to simple, sinister

And he points out something that I hadn't entirely internalized about
Qanon: it's a literal death cult. The core tenets of its adherents
include the idea that their enemies should be mass-murdered. Seen in
that light, it's no wonder than Q stuff is so creepy.

76 minutes' worth of video is a big ask, and to be honest, I listened to
(rather than watched) most of this (it still worked). Despite the
length, I'd recommend it. Olson's made a real contribution to the
discourse around conspiracism and the digital world.


👤 A cryptographic mystery solved

Andrey Bezrukov and Elena Vavilova were Russian spies who operated in
the USA for 20 years (this is the basis for "The Americans"); they were
caught in 2010. "Compromised," is the new memoir by Peter Strzok, the
FBI agent who had their case.


As Matt Blaze writes, a throwaway detail in the book resolves a
longstanding cryptographic mystery: that of a Cuban "numbers station"
that operated for years, including a decade where it behaved very
erratically (by numbers station standards).


Some background. Numbers stations - ratio stations in which people (or
synthesized voices) read out strings of random numbers - are a means of
messages for use with "one-time pads," a cryptographic tool that is, in
theory, unbreakable.

One-time pads are collections of random numbers used to encipher
messages through simple operations: adding each byte of your message to
the next number on the pad. If the pad is truly random,  secret and
never reused, the code can't be broken.


If your spies are sent abroad with a thick one-time pad, then you can
simply broadcast your messages over the entire region in which they
operate, and they can use their pads to decipher the messages, while
your adversaries just get random numbers.

Numbers stations, like the powerful shortwave transmitter in Bauta,
Cuba, were used to communicate with Soviet (and, later, Russian) spies
in the US in this way.

Though one-time pad messages can't be deciphered, it's still possible to
leak information using numbers stations. If a radio station ceases
operation every time a spy travels, then your adversary can match the
station's operating schedule with suspects' itineraries.

To prevent this "traffic analysis" attack, the station broadcasted dummy
traffic (random numbers that *weren't* encoded messages) every single
day, even if the spies were not listening that day.

However, for mysterious reasons - still not understood - the dummy
traffic never contained the number nine ("nueve"). That made it easy to
tell the real numbers station traffic from the dummy traffic, and from
there, it was possible to derive the spies' travel schedules.

Even with this glaring error, it was a *decade* before the FBI made
their their move. That was a whole decade in which the Cuban numbers
station was making this weird, stupid blunder.

One-time pads are incredibly powerful, but they're also super-awkward
and unforgiving. An error as simple as pad re-use can blow them up, as
happened with the notorious Venona affair:


As Blaze writes, "OTPs have long been a favorite of hucksters selling
supposedly 'unbreakable' crypto. Remember this story next time someone
tries to sell you their super-secure one-time-pad crypto. If actual
Russian spies can't use it securely, chances are neither can you."

Blaze was one of the researchers who followed - and recorded! - the
Cuban numbers station, and noted the mysterious and telling absence of
"nueve" in some of the traffic. He's posted a recording of the station
to his site:



👤 Youtube's war on algorithmic radicalization

Writing for Wired, Clive Thompson gets a first-of-its-kind
behind-the-scenes look at Youtube's algorithm development team, in order
to document the company's attempt to reduce the service's role in
spreading and reinforcing conspiracy theories.


Thompson traces the origin of the crisis to the company's drive for more
"engagement" that led them to tune their recommendation system to
identify and propose more specialized, esoteric versions of the video
you'd just watched.

The idea was to lead you down a rabbit hole of ever-more-specific
versions of your interests, helping you discover niches you never knew

This dynamic in recommendation systems has gotten a lot of attention
lately, and most of it is negative, but let's pause for a moment and
talk about what this means for non-conspiratorial beliefs.

Say you happen upon a woodworking video, maybe due to a friend's post on
social media. You watch a few of them and you find yourself interested
in the subject and tuning in more often.


The recommendation system presents an array of possible next-views, but
tilted away from general-interest woodworking videos, instead offering
you a menu of specialized woodworking styles, like Japanese woodworking.


You sample one of these and find it fascinating, so you start watching
more of these. The recommendation system clues you in to Japanese
nail-free joinery:


And from there, you discover the frankly mesmerizing
"Niju-mizu-kumi-tsugi" style of joinery, and you start to seek out more.
You have found this narrow, weird, self-reinforcing community.


This community could not exist without the internet and its signature
power to locate and connect people with shared, widely dispersed,
uncommon interests.

This power isn't just used to push conspiracies and woodworking
techniques, either.

It's how people who know that their gender identity doesn't correspond
to the gender they were assigned at birth find each other, and acquire a
vocabulary for describing their views, and foment change.

It's how people who believe Black Lives Matter find one another, it's
how the Green New Deal movement came together.

It's also how people who wanted to cosplay Civil War soldiers in
Charlottesville, waving tiki torches, chanting "Jews will not replace
us" found each other.

And that is the conundrum of the recommendation engine. Helping people
find others who share their views, passions and concerns is not, in and
of itself, bad. It is vital. It's the thing that made the internet
delightfully weird. It's also what made the world terribly weird.

Thompson takes us inside Youtube's algorithm team as they try to balance
three priorities:

I. Increasing their traffic and profits

II. Helping people find others with common interest

III. Stopping conspiracies from spreading

And he traces how they try, with limited success, to manage these
competing goals by creating extremely fine-grained rules that define
what is banned on the platform.

But naturally, this just gives rise to a new kind of content: stuff that
is ALMOST bad enough for blocking, but not quite. The problem is that
this stuff is indistinguishable (in all but the narrowest, technical
way)  from banned content.

So then Youtube has to create a new set of moderation guidelines: "What
is so close to prohibited content that it, too, is prohibited?"

Naturally, this is creating a new kind of content: "Stuff that is not
close-to-bannable, but is close-to-close-to-bannable."

This dynamic should be familiar to anyone who's watched the moderation
policies of Big Tech platforms evolve: what is "hate speech?" "What is
'almost-hate-speech?'" "What is almost-almost-hate-speech?'"

Ultimately, this ends up creating thick binders of pseudo-law that
delivers advantages to the worst people: they can study the companies'
policies and figure out how to skate *right up* to the cliff's edge (no
matter how it is defined).

And at the same time, they can goad their adversaries - the people they
torment - into crossing these fractally complex lines and then nark them
out, so that over time, these speech policies preferentially block good
speech and leave bad speech untouched.

I am increasingly convinced that the problem isn't that Youtube is
unsuited to moderating the video choices of a billion users - it's that
no one is suited to this challenge.

Remedies that put moderation choices closer to the user - breaking up
monopolies, allowing interoperable recommendation systems - solve the
problem of scaling up *and* covering edge cases by eliminating scale
altogether, and letting the edge cases make their own calls.


👤 The Hardware Lottery

Google Brain researcher Sara Hooker's new paper "The Hardware Lottery,"
proposes that there is a secret, relentless force that bends the course
of machine learning: processor architecture.


Hooker observes that computer scientists are curiously indifferent to
the constraints and capabilities of processors, and proceed in a vacuum
of information about these, designing machine learning algorithms that
are sometimes stymied by bottlenecks in processor designs.

While other machine learning techniques thrive and grow to dominate, not
merely because they do something useful, but because they do something
useful that can be readily accomplished with the hardware that we can
currently access.

(This indifference is multidirectional: hardware researchers often
ignore systems and algorithm designers, while systems designers ignore
hardware and algorithms)

There is a LOT to this: computer science bends towards the
accomplishable: things that are cheaper to do in hardware happen more. I
think the rise of cryptocurrency, VR, and machine learning is connected
to the rise of cheap, massively parallel GPUs:


One of the very best people on this is Herb Sutter, whose classic
"Welcome to the Jungle" identifies the ways in which processors are
becoming increasingly specialized and combined in a system system ("A
Heterogeneous Supercomputer in Every Pocket").


Likewise important is Neil "Fab Lab" Gershenfeld's mind-blowing talk
about "nonbinary computing" from Shmoocon 2016:


The idea that available processor architectures exert both subtle and
overt selective pressure on research agendas is really important and

What's interesting about Hooker's paper is that she also explores how
this dynamic plays out from the hardware and systems side, where they
seem to keep getting caught flatfooted by algorithmic advances that
require them to retool.

Disciplinary specialization has allowed systems, algorithms and hardware
research to make huge advances, but it has also reached a breaking point
due to poor coordination.

Meanwhile, generalization in processor and system design has ALSO been
tapped dry, necessitating a switch to specialized designs purpose-built
for different tasks.

There's something really magic about watching the whole field stumble
due to BOTH generalization AND specialization.


👤 This day in history

#5yrsago Tell-all free-to-play-game dev's confessions

#5yrsago Poker malware infects your computers and peeks at your cards


👤 Colophon

Today's top sources: Kottke (https://kottke.org/), Bruce Schneier
(https://schneier.com/), Four Short Links

Currently writing: My next novel, "The Lost Cause," a post-GND novel
about truth and reconciliation. Yesterday's progress: 517 words (62625

Currently reading: Gideon the Ninth, Tamsyn Muir

Latest podcast: IP https://craphound.com/podcast/2020/09/14/ip/

Upcoming appearances:

* Keynote for Law Via the Internet conference, Sept 22,

* DWeb Meetup — If Big Tech Is Toxic, How Do We Build Something Better?
Sept 22,

* Writing into an Uncertain Future, Afterwords Festival, Oct 1,

Latest book:

* "How to Destroy Surveillance Capitalism": an anti-monopoly pamphlet
analyzing the true harms of surveillance capitalism and proposing a

* "Little Brother/Homeland": A reissue omnibus edition with a new
introduction by Edward Snowden:
https://us.macmillan.com/books/9781250774583; personalized/signed copies

* "Poesy the Monster Slayer" a picture book about monsters, bedtime,
gender, and kicking ass. Order here:
https://us.macmillan.com/books/9781626723627. Get a personalized, signed
copy here:

Upcoming books:

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

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 pluralistic.net.


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


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


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

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: OpenPGP digital signature
URL: <http://mail.flarn.com/pipermail/plura-list/attachments/20200918/35f3a991/attachment.sig>

More information about the Plura-list mailing list