Start Up No.1783: US sanctions Russian bitcoin miners, the poetry of repetition evasion, why online offence is inevitable, and more

The UK government is proposing to allow self-driving cars to do limited driving on motorways. Odd, because cars with that capability are already here and driving around. CC-licensed photo by Edsel Little 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 9 links for you. Not on autopilot. I’m @charlesarthur on Twitter. Observations and links welcome.

Self-driving car users could watch films on motorway under new DfT proposals • The Guardian

Tobi Thomas:


Users of self-driving cars will be able to watch films on the motorway under planned changes to the Highway Code, although it will remain illegal to use mobile phones.

The update, proposed by the Department for Transport (DfT), will allow those in the driver’s seat to use a car’s built-in screens to watch movies and TV programmes.

The new rules also state that insurance companies will be financially liable, rather than individual motorists, for accidents in self-driving cars.

However, those behind the wheel must be ready to resume control of the vehicle when they are prompted – such as when they approach motorway exits. These measures were described as an interim measure by the government to support the early deployment of self-driving vehicles.

Although there are no vehicles currently approved for self-driving on roads in the UK, the first could be approved later this year. The introduction of the technology is likely to begin with vehicles travelling at slow speeds on motorways, such as in congested traffic.

In April 2021, the DfT said it would allow hands-free driving in vehicles with lane-keeping technology on congested motorways. Existing technology, including cruise control and automatic stop/start, is classified as being “assistive”, meaning that users must remain fully in control.


This is weird. Tesla cars (and others) can already do automatic lane-keeping with adaptive cruise control so they stay a specified time/distance from the car in front. They’ll do it at a lot faster than 37mph too. And can you watch TV on your mobile phone? If not, why not?

Tesla meanwhile has more than 100,000 people trying its “Full Self-Driving” beta program. Wonder if any of them encounters roundabouts – surely the last thing that FSD will be able to solve, given the ambiguity of intention and subtle dance it implies.
unique link to this extract

The Twitter account that collects awkward, amusing writing • The New Yorker

Naaman Zhou:


I first encountered Second Mentions in 2017, when I was working as a reporter in Sydney, Australia. News writing, by necessity, brings you up against repetition as an occupational hazard, and the account was shown to me by an editor, who shared it with a self-aware, nerdy, professional glee. The account tracks the ways that writers strive to express the same thing differently, with examples taken mostly from newspapers and magazines around the world. (A “second mention”—also known as a second reference—is the account’s name for these ways of avoiding repetition.) Take, for example, Adele, who is frequently “the singer Adele” on first mention, and then maybe “the Tottenham soul-pop titan” on second mention. Cheese, if you are saying “cheese” too much, can be “the popular dairy product.” A “pair of armadillos,” who, for some reason, were put on a diet? “The oval-shaped duo.” The account is addictively funny, and its discoveries are—variously—charming, insane, perfect.

Some greatest hits: the Times of London describing “tea” as “the bitter brown infusion.” The Guardian describing a fox who ran onto a soccer field as “the four-legged interloper.” The New York Times describing Grumpy Cat, the Internet meme, as “the sourpuss with the piercing look of contempt.” (In the cat’s obituary, no less.) Even this magazine, last year, describing electric scooters as “the long-necked, flat-bottomed machines.”


Though Zhou points to a real reason why one avoids repetition, as a journalist one just absorbs the necessity to do so through a sort of osmosis. And then you turn out either to be good or bad at it.
unique link to this extract

US Treasury sanctions Russian bitcoin miners • CNBC

MacKenzie Sigalos:


According to data from Cambridge University, Russia is the world’s third-biggest destination for bitcoin mining.

“By operating vast server farms that sell virtual currency mining capacity internationally, these companies help Russia monetize its natural resources,” Under Secretary for Terrorism and Financial Intelligence Brian Nelson said in a news release released early Wednesday afternoon.

“Russia has a comparative advantage in crypto mining due to energy resources and a cold climate. However, mining companies rely on imported computer equipment and fiat payments, which makes them vulnerable to sanctions,” continued the statement.

The US views income from the crypto mining industry as a potential threat to the efficacy of its sanctions regime, with the Treasury saying that it is committed to ensuring that no asset becomes a mechanism for the Putin regime to offset the impact of sanctions.

Among the companies targeted by US sanctions is BitRiver, which was founded in 2017, and as the name implies, operates its mining farms with hydroelectric power. The mining firm employs over 200 full-time staff in three offices across Russia, according to its website.

The Office of Foreign Assets Control has singled out 10 Russia-based subsidiaries of BitRiver in its most recent raft of sanctions on businesses and individuals helping Russia soften the blow of economic penalties.

…The IMF notes that the share of mining in sanctioned countries is “relatively contained.” It estimates that the monthly average of all bitcoin mining revenues last year was about $1.4bn, of which Russian miners could have captured close to 11% and Iranian miners 3%.


If – big if – the US can block exports of GPUs and other computing kit to BitRiver, it might have an impact in six months or so.
unique link to this extract

Energy Vault loses $1.2bn/40% market cap, CO2e/kWh worse than natural gas • CleanTechnica

Michael Barnard:


while I haven’t specifically torn apart Carbon Capture, it’s just another direct air capture solution, requiring absurd amounts of manufactured materials and energy to separate 415 ppm of CO2 from the air. As I calculated while assessing Carbon Engineering, an alternative technology which is also funded by the crowd around Bill Gates, you have to filter a Houston Astrodome’s worth of air to get a ton of CO2. Similarly, all the air in the Grand Canyon contains only 1,270 tons of CO2. The Canyon’s volume is 1.67 billion cubic metres.

Getting up a to a million tons of CO2 a year would require two kilometers of 20-metre high, 8-metre thick fans running 24/7/365 for 0.0025% of annual CO2 emissions, and 0.0001% of the historical problem. As this isn’t about Carbon Capture, I’ll just say that Carbon Engineering’s solution is only fit for enhanced oil recovery using unmarketable natural gas, which is exactly what it’s doing in the Permian Basin with Oxy. The natural markets for direct air capture are capturing governmental funding, oil and gas greenwashing, and enhanced oil recovery, none of which merit investment in 2022.

And now there is Energy Vault, making Gross’ contribution to climate solutions a trifecta of challenges.

The initial concept was terribly silly in obvious ways, which didn’t prevent a lot of money from being thrown at it. It involved cranes picking up big concrete blocks and stacking them in an increasingly high circle around the cranes to store energy and lowering them back down to the ground again to release energy. It was the concept and prototype I first looked at and then ignored as it wasn’t worth my time to debunk it.


Now he does go on to debunk it. Gravitational storage, using blocks of objects rather than water, turns out to be a terrible, bad, no good idea.

Useful to see how much air you need to process to capture CO2. This makes it clear that “carbon capture” using anything other than natural processes, or some amazing chemistry, isn’t going to work.
unique link to this extract

You can’t police offence • UnHerd

Kathleen Stock, who gets a fair dollop of online (and offline) abuse, isn’t impressed by the proposals in the Online Harms Bill:


Following scrutiny from the Joint Committee, the Bill — which received its second reading in the Commons yesterday — takes recent Law Commission proposals to introduce a “harm-based” communications offence, and places a duty of care on internet providers and websites to restrict content which meets the definition of this proposed offence, give or take a few tweaks. Specifically, they will be required to restrict any content where there’s a “real and substantial risk that it would cause harm to a likely audience”, the sender “intended to cause harm to a likely audience”, and the sender has “no reasonable excuse for sending the message”. Harm is defined as “psychological harm amounting to at least serious distress”. What counts as a “likely audience” comprises whichever individual is reasonably foreseen as encountering that content.

The flaws here were also present within the Law Commission proposals that inspired the Bill. Take the criterion of “psychological harm amounting to at least serious distress”. As many have noted — though apparently not in Westminster — concepts such as “psychological harm” and “distress” are moving targets, semantically speaking, in the sense that the sort of thing they refer to changes over time. For instance, in a society whose primary concern is with the alleviation of negative experience, concepts associated with negative experiences tend to expand their semantic range and become increasingly diluted. So for instance, over time, the category of “abuse” has moved beyond physical events to include emotional ones as well; and the category of “trauma” has extended from atypically catastrophic life events to relatively common happenings like childbirth and bereavement.

At first glance, “experts discover new form of trauma!” looks reassuringly scientific, a bit like “experts discover new kind of dinosaur!”. But whereas the existence of a dinosaur is completely independent of the activities of the experts who discover it, this is not the case with trauma.


On the day this appeared, a man was given a suspended 10-week jail sentence for sharing a video in 2018 of a bonfire in which he’d burnt a model of Grenfell Tower (which caught fire, killing multiple people, in 2017) and made remarks like “that’s what happens when you don’t pay the rent”.

Unfunny? Sure. Offensive? Sure. Worthy of a prison sentence? Not by any reasonable definition. He was acquitted once, and the government appealed and that was quashed. Then they tried again.
unique link to this extract

How DALL-E 2 actually works • AssemblyAI

Ryan O’Connor is the Developer Educator at OpenAI:


OpenAI’s groundbreaking model DALL-E 2 hit the scene at the beginning of the month, setting a new bar for image generation and manipulation. With only short text prompt, DALL-E 2 can generate completely new images that combine distinct and unrelated objects in semantically plausible ways, like the images below which were generated by entering the prompt “a bowl of soup that is a portal to another dimension as digital art”.

DALL-E 2 can even modify existing images, create variations of images that maintain their salient features, and interpolate between two input images. DALL-E 2’s impressive results have many wondering exactly how such a powerful model works under the hood.

In this article, we will take an in-depth look at how DALL-E 2 manages to create such astounding images like those above. Plenty of background information will be given and the explanation levels will run the gamut, so this article is suitable for readers at several levels of Machine Learning experience. Let’s dive in!


The water’s very deep – it includes a concept called a “diffusion model” which is “a thermodynamics-inspired invention” which learns to generate data by reversing a gradual noising process. Got it?

But DALL-E 2 (as in WALL-E, but for Drawing) has impressed lots of people.
unique link to this extract

Brave’s De-AMP feature bypasses ‘harmful’ Google AMP pages • The Verge

David Pierce:


Brave announced a new feature for its browser on Tuesday: De-AMP, which automatically jumps past any page rendered with Google’s Accelerated Mobile Pages framework and instead takes users straight to the original website. “Where possible, De-AMP will rewrite links and URLs to prevent users from visiting AMP pages altogether,” Brave said in a blog post. “And in cases where that is not possible, Brave will watch as pages are being fetched and redirect users away from AMP pages before the page is even rendered, preventing AMP / Google code from being loaded and executed.”

Brave framed De-AMP as a privacy feature and didn’t mince words about its stance toward Google’s version of the web. “In practice, AMP is harmful to users and to the Web at large,” Brave’s blog post said, before explaining that AMP gives Google even more knowledge of users’ browsing habits, confuses users, and can often be slower than normal web pages. And it warned that the next version of AMP — so far just called AMP 2.0 — will be even worse.

Brave’s stance is a particularly strong one, but the tide has turned hard against AMP over the last couple of years.


Not going to have a measurable effect, but as a PR gesture it gets great visibility.
unique link to this extract



How to play:

🕵️‍♂️ You have 60 seconds to solve all 10 clues

1️⃣ Change 1 letter from the previous word to get the answer

🙅‍♀️ You are not penalized for incorrect guesses

⏭ You have 1 skip (costs 5 seconds)

🕛 New puzzle at midnight


So FANG is followed by BANG is followed by BANK is followed by TANK etc. It’s quite fun, if your daily Wordle leaves you a little short.
unique link to this extract

Netflix explores a version with ads as subscriber base shrinks • WSJ

Joe Flint and Denny Jacob:


The move is a significant change for a company that has sold itself since its inception as a commercial-free haven for its members. Netflix is grappling with slowing revenue growth caused by stiffer competition from rival services and rampant account sharing among its customers.

In a Tuesday analyst interview to discuss the company’s first quarter-results, Netflix chairman and co-chief executive Reed Hastings said an ad-supported version of Netflix makes a lot of sense.

“Those who have followed Netflix know that I’ve been against the complexity of advertising and a big fan of the simplicity of subscription,” Mr. Hastings said. “But as much as I’m a fan of that, I’m a bigger fan of consumer choice.”

Netflix earlier in the day said it ended the first quarter with 200,000 fewer subscribers than it had in the fourth, missing on its own projection of adding 2.5 million customers in the period. Netflix said it expected to lose two million global subscribers in the current quarter.

Netflix shares were 25% lower in after-hours trading. Through Tuesday’s close, the stock has declined by more than 40% so far this year.

…Netflix warned that gains made during the Covid-19 pandemic hid the fault lines that have emerged in its business over the past few years. “Covid clouded the picture by significantly increasing our growth in 2020, leading us to believe that most of our slowing growth in 2021 was due to the Covid pull forward,” the company said in its letter.

…Although Netflix has several hit shows including “Stranger Things,” “Bridgerton” and “The Crown,” the service has also had its fair share of expensive flops recently including “Jupiter’s Legacy” and “Space Force.”

“We need to have an ‘Adam Project’ and a ‘Bridgerton’ every month,” said Co-Chief Executive and Chief Content Officer Ted Sarandos during the analyst interview.


The question is whether the existence of a (lower) tier that has ads would lead people to downgrade (and would that be revenue-neutral?) or would sully the brand, which has been ad-free, in contrast to US TV.
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?

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

Errata, corrigenda and ai no corrida: The Ethernet wiring you want for properly high-speed internet is Cat6a, not 6e (thanks @Cinemaworks).

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.