Start Up No.1747: Apple halts Russian sales, Russians buy crypto, the missing cyberwar, “yes, Putin would 🍄”, convoy tracking, and more

Machine learning still struggles to make good film recommendations. Why is that, though? CC-licensed photo by Elias Bizannes on Flickr.

You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.

A selection of 10 links for you. Still mostly Ukraine. I’m @charlesarthur on Twitter. Observations and links welcome.

Apple halts all sales from online store and to channels in Russia • MacRumors

Juli Clover:


Apple today confirmed that it has stopped all product sales from its online website in Russia, which means customers in Russia can no longer purchase Macs, iPhones, iPads, and other Apple devices. Attempting to make a purchase from the Russia store results in a “delivery unavailable” result when trying to add a product to the online cart.

Sales have been halted following a plea last week from Ukrainian vice prime minister Mykhailo Fedorov, who wrote a letter to Apple CEO Tim Cook asking Apple to stop device sales and to block App Store access in Russia.


I appeal to you and I am sure you will not only hear, but also do everything possible to protect Ukraine, Europe, and finally, the entire democratic world from bloody authoritarian aggression – to stop suppling Apple services and products to the Russian Federation, including blocking access to App Store!

We are sure that such actions will motivate youth and active population of Russia to proactively stop the disgraceful military aggression.


Apple said in a statement that it has also stopped all exports into the sales channel in the country and disabled traffic and live incidents in Apple Maps in Ukraine as a safety and precautionary measure for Ukrainian citizens.


A lot easier for Apple to do, as has been pointed out by many, when it doesn’t have substantial manufacturing in Russia and it doesn’t represent 20% of its sales.
unique link to this extract

Russians are buying more crypto as sanctions set in, data shows • Vice

Ekin Genç:


The Russian ruble fell to a record low against the US dollar on Monday, plummeting by 30% at one point before regaining a third of its losses as the Central Bank of Russia hiked the interest rates from 9.5% to 20%. High interest rates often lure local currency depositors in a bid to help stop further depreciation, or at least that’s the goal.

Now, there are indications that Russians are converting their rubles into cryptocurrency in a bid to protect the value of their savings. Specifically, much of the buying activity has centered on Tether (USDT), a stablecoin that is pegged 1:1 with the value of the US dollar.

Data from blockchain research firm Arcane Research shared with Motherboard shows that USDT/RUB (Tether/Russian ruble) trading volume on Monday broke a new record with $34.94m. The previous daily record, $34.31m, was in May of last year, when Bitcoin’s price came crashing down after Elon Musk criticized its environmental footprint, and many investors—not just Russians—switched to stablecoins.

Monday’s trading volume in Tether was 519% above the average for this year—a period when Russian invasion rumors and possible sanctions were already circulating.

Last week, Bitcoin saw a 214% week-over-week growth in BTC/RUB volume, compared to a 46% growth in the global volumes in the same period, Arcane analyst Vetle Lunde told Motherboard. “Russian investors are evidently far more active in the market compared to the global investors,” he said.

The ruble-denominated Bitcoin trade recorded $11.4m in daily volume on Monday—a large amount but a far cry from the stablecoin trade volume in rubles.


Honestly, that’s small beer – and Tether implies it would take action if people tried to move very large amounts of money through these exchanges. This Twitter thread (on a single page) also explains why it’s not going to be a way to evade sanctions.
unique link to this extract

Russia, Ukraine cyberwar hasn’t unfolded as expected • The Washington Post

Joseph Menn and Craig Timberg:


Ukraine’s core cyberdefense has done better than expected because it focused on the issue after Russian hackers briefly knocked out power to swaths of the country in 2015 and 2016, said David Cowan, a veteran cybersecurity venture capitalist and corporate director, and because it has had help from American and European experts.

“I would have thought that by now Russia would have disabled a lot more infrastructure around communications, power and water,” Cowan said. “If Russia were attacking the US, there would be more cyber damage.”

The absence of major disruptions predicted by cyberwar doctrine has allowed Ukraine’s President Volodymyr Zelensky to deliver propaganda coups with little more than a smartphone and a data link. Images of civilian casualties, the brutal shelling of cities and also some Russian losses have undermined that nation’s claims of a limited and humane “special military operation.” A viral audio clip of Ukrainian soldiers on a tiny island telling a Russian warship to “go f— yourself” has become a defining moment of national resistance.

“It’s become a global participatory thing. Everybody thinks they’re part of it,” said Doug Madory, director of Internet analysis for Kentik, which tracks global data flows. “It would be a lot harder to do all that if there was a blackout.”

Ukraine has not escaped unscathed, and some experts warn that cyberattacks or Internet outages could grow as Russia’s invasion intensifies in the face of unexpectedly stout resistance.

Russia or its allies already have deployed software to wipe data off some Ukrainian computers, including border control offices. But such intrusions are not nearly as widespread as in past attacks such as NotPetya, in which fake ransomware attributed to the Russian government caused billions of dollars in damages, much of it in Ukraine.


When the story of this invasion comes to be written (there’s implicit optimism for you) the failure to knock out the mobile networks will look like a key mistake. (Though there is wiper software at large.)
unique link to this extract

‘Yes, he would’: Fiona Hill on Putin and nukes • POLITICO

Maura Reynolds:


Hill spent many years studying history, and in our conversation, she repeatedly traced how long arcs and trends of European history are converging on Ukraine right now. We are already, she said, in the middle of a third World War, whether we’ve fully grasped it or not.

“Sadly, we are treading back through old historical patterns that we said that we would never permit to happen again,” Hill told me.

Reynolds: What have we learned about NATO in the last two months?

Hill: In many respects, not good things, initially. Although now we see a significant rallying of the political and diplomatic forces, serious consultations and a spur to action in response to bolster NATO’s military defenses.

But we also need to think about it this way. We have had a long-term policy failure going back to the end of the Cold War in terms of thinking about how to manage NATO’s relations with Russia to minimize risk. NATO is a like a massive insurer, a protector of national security for Europe and the United States. After the end of the Cold War, we still thought that we had the best insurance for the hazards we could face — flood, fire etc. — but for a discounted premium. We didn’t take adequate steps to address and reduce the various risks. We can now see that that we didn’t do our due diligence and fully consider all the possible contingencies, including how we would mitigate Russia’s negative response to successive expansions. Think about Swiss Re or AIG or Lloyds of London — when the hazard was massive, like during Hurricane Katrina or the global financial crisis in 2008, those insurance companies got into major trouble. They and their clients found themselves underwater. And this is kind of what NATO members are learning now.

…if anybody thinks that Putin wouldn’t use something that he’s got that is unusual and cruel, think again. Every time you think, “No, he wouldn’t, would he?” Well, yes, he would. And he wants us to know that, of course.


She also calls for a complete temporary suspension of business activity with Russia: that companies have a choice to make about whether they want to help a regime doing this. An essential read. The money quote: “We’re already in World War III.”
unique link to this extract

Conti ransomware’s internal chats leaked after siding with Russia • BleepingComputer

Lawrence Abrams:


A Ukrainian security researcher has leaked over 60,000 internal messages belonging to the Conti ransomware operation after the gang sided with Russia over the invasion of Ukraine.

BleepingComputer has independently confirmed the validity of these messages from internal conversations previously shared with BleepingComputer regarding Conti’s attack on Shutterfly.

AdvIntel CEO Vitali Kremez, who has been tracking the Conti/TrickBot operation over the last couple of years, also confirmed to BleepingComputer that the leaked messages are valid and were taken from a log server for the Jabber communication system used by the ransomware gang.

Kremez told BleepingComputer that the data was leaked by a researcher who had access to the “ejabberd database” backend for Conti’s XMPP chat server. This was also confirmed by cybersecurity firm Hold Security.

In total, there are 393 leaked JSON files containing a total of 60,694 messages since January 21, 2021, through today. Conti launched their operation in July 2020, so while it contains a big chunk of their internal conversations, it is not all of them.

These conversations contain various information about the gang’s activities, including previously unreported victims, private data leak URLs, bitcoin addresses, and discussions about their operations.


Initially it was suggested this happened because one of the gang is Ukrainian and didn’t like the Russian invasion, but it seems just to have been an independent security researcher. The bitcoin wallets associated with the group have apparently received hundreds of millions of dollars in payments. Now they’re screwed for a couple of months at least.
unique link to this extract

The speed of information • SatPost

Trung Phan:


Twitter was founded in 2006. The most recent conflict of this *size* was the Iraq War, which started in 2003.

Twitter is already optimized as a dopamine drip machine. Now, it’s covering the largest land invasion on the European continent since the end of World War II. Then layer on the drama of a David vs. a nuclear-armed Goliath battle (Zelensky + Ukraine vs. Putin + Russia). Finally, throw in confusion as to what information is real and it has truly become “insane”.

Like a lot of writers I know, I have dozens of Google docs with ideas that “I might write about one day”. A random topic I’ve been collecting notes on is “what was communication like back in the day”.

I started this vaguely titled list after reading about how Abraham Lincoln received information during the US Civil War. After the war began in April 1861, the US Military Telegraph Corps. laid “15,000 miles of telegraph wire across battlefields that transmitted news nearly instantaneously from the front lines”, per History.

All communications from that telegraph network — literally 100% — was sent to the library room of the War Department, which was next to the White House.

Other than the White House, Lincoln spent more time in the telegraph room than any other place during the Civil War:


David Homer Bates, one of the four original members of the U.S. Military Telegraph Corps, recounted in “Lincoln in the Telegraph Room” that several times a day, Lincoln sat down at a telegraph office desk near a window overlooking Pennsylvania Avenue and read through the fresh stack of incoming telegrams, which he called “lightning messages.” As telegraph keys chattered, he peered over the shoulders of the operators who scribbled down the incoming messages converted from Morse Code. He visited the office nearly every night before turning in and slept there on a cot during pivotal battles.


From 1861-1865, the President of the United States was the only person in the country receiving all the flow of information related to the war (the Confederates never built a comparable telegraph network). He wrote more than a thousand telegrams.

Today, literally billions of people are being flooded with images, intel, news, updates and propagandas at every waking second. Obviously, we’re not getting the full picture but it’s an astounding amount of information (and mis/dis-information, too).

Each of us has Lincoln’s telegraph room in our pocket.


Plenty more analysis in the post, but even this part is quite astonishing.
unique link to this extract

Facebook AI researchers built a ‘fashion map’ with your social media photos • Vice

Ella Fassler:


Artificial intelligence researchers—some of whom are affiliated with Facebook’s parent company Meta and Cornell University—used more than 7 million public, geolocated social media photos from Instagram and Flickr to construct what they’re calling an “underground fashion map” that spans 37 cities. The map can reveal groupings of people within a city, including areas that are the most “trendy” or “progressive,” and builds on an Amazon-funded AI tool called GeoStyle to forecast fashion trends, according to a press release about the research.

“A person unfamiliar with a city could find out what neighborhoods might be suitable for them to visit, e.g., to satisfy interests in outdoor activities vs. shopping vs. tourist areas,” researchers wrote in a newly published report completed as part of an internship with Facebook AI Research. They also claim anthropologists could leverage the maps to infer trends within a city across time.

The project’s affiliation with Facebook and Amazon raises larger questions about the unexpected ways tech companies use personal data, often without explicitly notifying users.

Tamara Berg, co-author of the report and director of Meta AI—Facebook’s artificial intelligence research center—did not respond to Motherboard’s inquiry about Facebook’s potential use of the data, or whether Instagram and Flickr users are aware that their photos are being used to construct fashion maps.


Everything old is new again: back in January 2014 I wrote about a little British company called Jetpac, which produced “city guides” based on content from Flickr and Instagram. Its co-founder and chief technology officer Pete Warden was hired by Google, which was rather taken by the idea. But of course Instagram is owned by Facebook. And good ideas never die.
unique link to this extract

The oddly addictive quality of Google Alerts • The New Yorker

Casey Cep:


Google Alerts can cast a wonderful net, but mesh size matters: large holes and it catches nothing, too small and it catches everything. Consider the earliest and one of the most persistent reasons for setting these alerts: tracking yourself. All is vanity, perhaps especially on the Internet, so it’s no surprise that one of the things that we’re most eager to know is what the world is saying about us.

The engineer who developed the alert system for Google told CNN in 2016 that when he first presented the idea, twenty years ago, his manager was skeptical, worrying that it would starve the search-engine of traffic: rather than consumers constantly searching for fresh mentions of whatever topic interested them, they would wait for the alert, then follow its links not to Google but to outside Web sites, leaching away potential advertising revenue. In response, the engineer, one of the first forty or so employees of the company, took his prototype to Google’s co-founders, who approved it after watching him demonstrate only two search terms: “Google” and “Larry Page,” the name of one of the co-founders.

Learning what other people thought about us used to take either a great deal of luck, like Tom Sawyer being mistaken for dead and then getting to eavesdrop on his own funeral, or a great deal of effort, like Harun al-Rashid, a caliph of the Abbasid dynasty, in the “Arabian Nights,” disguising himself in order to venture out into the streets and talk with his subjects candidly. But the Internet has made it easy—made it, in fact, almost unavoidable. The same Google Alert can make sure you know that your long-lost bunkmate from summer camp has mentioned you in an essay, that a friend of your deceased uncle has written a memoir of their time together in the Marines (including the care packages you sent them), and that the local newspaper has digitized its archives, thereby offering up to the internet your high-school football averages and your arrest for vandalism.


unique link to this extract

Kyiv Convoy Tracker

Corey Scher:


This map automatically updates to highlight diferences between Sentinel-1 radar images at the location of Russian convoy buildup north of Kyiv.


Another fascinating piece of open source intelligence work. At the time of checking (about 11pm Tues Kyiv time) the head of the convoy was about 57km to the north.
unique link to this extract

My favorite movies of 2021 • Remains of the Day

Eugene Wei:


Film remains a difficult category for machine learning to crack. Most people only watch movies once. In a category like music, people listen to their favorite tracks repeatedly. Films are very long while music tracks only last a few minutes. As a result, the frequency of feedback is much higher for music than film.

Viewers generally provide a single point of feedback on a film, if they even choose to sample it: they either finish the movie or they don’t. In music, you not only gather many more data points per hour because of the short duration of each track, but you gather feedback within each piece. People hit skip, or rewind, or repeat. People add songs to playlists or ask their streaming service to generate radio stations off of that track.

As I’ve written before about TikTok, one of its most critical design choices was to full-screen videos, allowing it to gather really accurate signal from the viewer on each video. TikTok videos are even shorter than music tracks, but they often contain snippets of music tracks. In many ways a TikTok is about as short a piece of media as could be designed that can be said to still tell a narrative (though maybe a dating app profile photo is even more concise).

The ways that music tracks resemble each other feel easier to see with math. This makes it easier to generate a playlist of similar tracks even before gathering listener feedback. Machine learning algorithms have learned to write music that often sounds like specific composer and musicians. I’ve yet to see an algorithm that can just spit out a Wes Anderson-like movie.

It’s no surprise to me that Netflix seems largely to have given up on much of the work that came out of the Netflix Prize and instead focuses on using the massive funnel of its above-the-fold home screen real estate to push its latest original production. I didn’t like Red Notice, but I can understand what types of metrics would lead Netflix to just splash it across every subscriber’s eyeballs.


I haven’t watched any of the movies he liked, and of the TV series have only watched Succession. Though I agree with his analysis. It’s how humans work, isn’t it?

And apparently pushing new content, rather than showing you the programme you’d like to continue watching, makes millions for streaming companies.
unique link to this extract

• Why do social networks drive us a little mad?
• Why does angry content seem to dominate what we see?
• How much of a role do algorithms play in affecting what we see and do online?
• What can we do about it?
• Did Facebook have any inkling of what was coming in Myanmar in 2016?

Social Warming, my latest book, and find answers – and more.

Errata, corrigenda and ai no corrida: none notified

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

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