Start Up No.1998: TV writers’s strike gripe, the real web history, Apple’s AI struggles, the stretch limo vanishes, Netflix’s lost users, and more

The proliferating buttons on the car dashboards of a decade or so ago went away.. and now they’re coming back. CC-licensed photo by Elizabeth 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.

When it’s Friday, there’ll be another post at the Social Warming Substack at about 0845 UK time.

A selection of 10 links for you. Not even looking! I’m @charlesarthur on Twitter. On Mastodon: Observations and links welcome.

The glorious return of a humble car feature • Slate

David Zipper:


As I explained in a 2021 Slate article, the trend toward car touch screens has been a dangerous one for road safety. Those who drove in the 1990s will remember using buttons and knobs to change the radio or adjust the air conditioning without looking down from the steering wheel. Despite their name, touch screens rely on a driver’s eyes as much as her fingers to navigate—and every second that she is looking at a screen is a second that she isn’t looking at the road ahead. Navigating through various levels of menus to reach a desired control can be particularly dangerous; one study by the AAA Foundation concluded that infotainment touch screens can distract a driver for up to 40 seconds, long enough to cover half a mile at 50 mph.

“The irony is that everyone basically accepts that it’s dangerous to use your phone while driving,” said Farah. “Yet no one complains about what we’re doing instead, which is fundamentally using an iPad while driving. If you’re paying between $40,000 and $300,000 for a car, you’re getting an iPad built onto the dashboard.”

Seeking to address these risks, NHTSA published voluntary guidance in 2013 recommending that a driver be able to complete any infotainment task with glances of under two seconds, totaling a maximum of 12 seconds. But NHTSA’s guidance had no enforcement mechanism, and carmakers have violated it with impunity.

In the last two years further evidence has suggested that touch screens represent a step backward for auto design. Drexel researchers found that infotainment systems posed a statistically significant crash risk even in the early 2010s, before carmakers added many of today’s bells and whistles. A widely publicized Swedish study found that completing tasks with screens takes longer than with physical buttons.

Meanwhile, a revolt has been brewing. A recent J.D. Power consumer survey on vehicle dependability concluded that “infotainment remains a significant issue for new vehicles.” It wasn’t hard to understand why. In a 2022 New York Times opinion piece titled “Touch Screens in Cars Solve a Problem We Didn’t Have,” Jay Caspian Kang wrote, “I can think of no better way of describing the frustration of the modern consumer than buying a car with a feature that makes you less safe, doesn’t improve your driving experience in any meaningful way, saves the manufacturer money and gets sold to you as some necessary advance in connectivity.”


Wonder if the NHTSA will mandate something to do with touchscreens, and how the carmakers – addicted to cheap screens – will cope with it.
unique link to this extract

Why are TV writers so miserable? • The New Yorker

Michael Schulman:


For people outside the industry, the woes of TV writers can elicit a boo-hoo response: it is, after all, a more lucrative form of writing than most, right? But the economics of streaming have chipped away at what was previously a route to a middle-class life, as the cost of living in Los Angeles has crept upward. “It feels like the studios have gone through our contracts and figured out how to Frankenstein every loophole into every deal, which means that, at the very best, you can keep your head above water,” [Laura] Jacqmin said. “You can maybe maintain the amount of money you made the year before, but more than likely you will be asked to cut your quote. It just feels really grim.” She added, “I’m on Twitter every other day, and I’m seeing writers who are, like, ‘Please Venmo me some grocery money. I am desperate, and I have not worked in three months. Help!’ ”

Aly Monroe, a 30-year-old writer who’d worked up from production assistant to story editor on Hulu’s “The Handmaid’s Tale,” told me that she makes about $10,000 a year in residuals, “and that’s certainly not reflective of what the studio is making.” In the long breaks between seasons, she relies on her wife’s more regular income while stretching out the money from “Handmaid.” Some of her friends are getting copywriting jobs or moving back in with their parents. “Before the strike demands came out, a lot of my friends were feeling really hopeless and essentially ready to give up, because it had just been such a hard road,” she said. “And they think that what the W.G.A. is asking for makes us all feel really good and like we’re working toward something that can make it back into a livable career for all of us. That’s certainly how I feel.”

At the same time that the money has tightened, original ideas have become harder to sell. The prestige-cable days of “Mad Men” and “Nurse Jackie” became the prestige-streaming era of “The Handmaid’s Tale” and “Stranger Things,” which has given way to the algorithm-and-I.P.-fuelled hellscape of superheroes, mergers, and HBO Max becoming plain old Max.


They’re also extremely worried about AI being used to generate ideas or content or to “punch up” content they’ve written. And given how studios will use absolutely any excuse to screw over those who work for them, it’s a legitimate concern.
unique link to this extract

Web3’s fake version of Web history • The Future, Now and Then

Dave Karpf:


Chris Dixon is a general partner at Andreessen Horowitz (a16z), one of the most influential Venture Capital firms in the world. He is Web3’s single biggest investor, and its most prominent evangelist.

And he is just atrocious at explaining the history of the Web.

Dixon made his money in the internet of the ‘00s and ‘10s. He works with Marc Andreessen, the iconic “golden geek” of the ‘90s internet. The guy has been around long enough to know better. When he gets the history of the Web completely wrong, he is doing so with intent. And he has been prominently, boldly getting this wrong for YEARS. No one seems to call him on it. I don’t understand why.

(Yes I do. It’s because he’s rich and well-connected. Picking fights with him over something like “the history of the web” has little upside. It’s one of those things that only a tenured professor who isn’t looking for much research funding would bother with.)

…Broad historical narratives are a bit like statistical models — “all models are wrong, but some models are useful.” Of course history is more messy and complicated than that. But if the general outline makes sense, and if it helps us make sense of the present, then the effort is justifiable.

But let me offer a corollary: “all models are wrong, but some are wronger than others.” And the problem with Dixon’s model is that it extremely, ceaselessly, aggressively wrong. It’s the type of wrong that might be useful for hawking unregistered Web3 security products (err, sorry, I mean, play-to-earn games), but is not at all useful for actually understanding the development of the internet.


Karpf offers a neat history of the internet – and it is correct, especially about the Web2 phase.
unique link to this extract

Report describes Apple’s “organizational dysfunction” and “lack of ambition” in AI • Ars Technica

Samuel Axon:


The Information’s sources [in an article about Apple’s struggles or otherwise with AI] offer up numerous examples of senior Apple leadership putting the brakes on (or at least reining in) aggressive efforts within the company’s AI group for fear of seeing products like Siri present the same kinds of embarrassing factual errors or unhinged behavior that ChatGPT and its ilk have done. In other words, Apple isn’t keen on tolerating what many working in AI research and product development call “hallucinations.”

For example, Siri’s responses are not generative—they’re human-written and human-curated. Apple leadership has been hesitant to allow Siri developers to push the voice assistant toward detailed back-and-forth conversations like you see in the latest LLM-driven chatbots. Those are seen as more attention-grabbing than usefulness, and Apple is worried about being responsible for bad answers.

Some engineers within the company have argued that Apple should be more tolerant of bizarre edge cases and factual errors, saying that a certain scale and comfort for wonkiness is needed to truly improve them. Notably, several senior people within the company have abandoned ship for Google or startups out of frustrations with Apple’s conservative mindset.

Further, Apple has increasingly focused on running AI and machine learning features on users’ local devices—both because that enabled faster response times and because of the company’s public commitment to user privacy. For some features, that is an advantage (as Giannandrea explained to Ars Technica in 2020). But to date, LLMs typically run in the cloud, and some have questioned whether they’ll ultimately work as well on local devices.


Counting down to LLMs on-device.. what do we think, a year?
unique link to this extract

The long demise of the stretch limousine • The New York Times

Jesus Jiménez:


Over a few days in early March, carmakers and limousine company operators gathered at the MGM Grand in Las Vegas for an annual convention, where they went to panels and parties and admired shiny new party buses, vans and black sport utility vehicles.

But something was missing.

“There wasn’t one stretch limousine on the show floor,” said Robert Alexander, president of the National Limousine Association, a trade group. “Not one.”

Decades ago, stretch limos were a symbol of affluence, used almost exclusively by the rich and famous. Over time, they became more of a common luxury, booked for children’s birthday parties or by teenagers heading to the prom.

These days, it seems as if hardly anyone is riding in a stretch limo. While the limousine name has stuck, the limo industry has shifted to chauffeur services in almost anything but actual stretch limos, which have largely been supplanted by black S.U.V.s, buses and vans.
“The limo business isn’t your father’s limo business anymore,” Mr. Alexander said.

Today, the stretch limo represents less than 1% of services offered by limo companies, down from about 10% a decade ago, according to the association.

“The stretch limo is — what’s the expression? — gone like the dodo bird,” Mr. Alexander said. “Extinct.”


What happened? Mostly, Uber and Lyft. The Great Recession. Also not looking like a jerk.
unique link to this extract

Netflix Spain lost 1 million users last quarter, Kantar says • Bloomberg via Yahoo

Clara Hernanz Lizarraga and Thomas Seal:


In early February, Netflix introduced a €5.99 ($6.57) monthly fee for users in Spain who shared their log-in details with another household and technical measures to detect such sharing. The move was linked to a fall in users of more than a million, two thirds of whom were using someone else’s password, according to Kantar’s research, which is based on surveys of household streaming habits.

“It’s clear this steep drop is due to the crackdown,” said Dominic Sunnebo, global insight director at Kantar’s Worldpanel Division, adding that the loss of a million users, even if most weren’t paid subscribers, would be a blow to Netflix in terms of word of mouth recommendation for its shows and service.

Subscription cancellations in the first quarter tripled compared to the previous period, according to Kantar’s research. Of all remaining Netflix subscribers in Spain, one-tenth said they planned to unsubscribe in the second quarter.

A similar fee was introduced in Portugal, Canada and New Zealand after a roll-out in several Latin American countries.

“We see a cancel reaction in each market when we announce the news,” Netflix said in its first quarter earnings release on April 18, expecting the dip to be momentary before users that didn’t pay start signing up for their own accounts.


Netflix says that Canada, where it tried this first, dipped and then came back above the previous point. Keep those fingers crossed!
unique link to this extract

Who Can I Vote For?


Find out about candidates in your area.

WhoCanIVoteFor is a simple tool which allows UK voters to see lists of candidates in upcoming elections using only their postcode.

All the candidate data ued in this site is collected by volunteers from council websites and other sources. If you wish to contribute, please visit Democracy Club Candidates.


OK, so this is a little late (today is the day for local elections in many, though not all, parts of the UK) but it’s a good site to bookmark – and also available in Welsh. Hats tipped to Sym Roe, Joe Mitchell, Tim Green, Andy Lulham and David Miller, with a little help from mySociety – the latter being one of those Web2 ideas that keeps on giving.
unique link to this extract

Amnesty International criticised for using AI-generated images • The Guardian

Luke Taylor:


While the systemic brutality used by Colombian police to quell national protests in 2021 was real and is well documented, photos recently used by Amnesty International to highlight the issue were not.

The international human rights advocacy group has come under fire for posting images generated by artificial intelligence in order to promote their reports on social media – and has since removed them.

The images, including one of a woman being dragged away by police officers, depict the scenes during protests that swept across Colombia in 2021.

But any more than a momentary glance at the images reveals that something is off.

The faces of the protesters and police are smoothed-off and warped, giving the image a dystopian aura.

The tricolour carried by the protester has the right colours – red, yellow and blue – but in the wrong order, and the police uniform is outdated.

…Amnesty International said it had used photographs in previous reports but chose to use the AI-generated images to protect protesters from possible state retribution.

To avoid misleading the public, the images included text stating that they were produced by AI.

“We have removed the images from social media posts, as we don’t want the criticism for the use of AI-generated images to distract from the core message in support of the victims and their calls for justice in Colombia,” Erika Guevara Rosas, director for Americas at Amnesty, said.

“But we do take the criticism seriously and want to continue the engagement to ensure we understand better the implications and our role to address the ethical dilemmas posed by the use of such technology.”


As Ryan Broderick pointed out in his Garbage Day email/Substack, Amnesty could completely have used real photos, and just blurred the faces of the people. But they thought to themselves it would be cool and edgy to use AI.
unique link to this extract

August 7 1998: US embassies in Africa bombed • BBC On This Day


At least 200 people have been killed and more than 1,000 injured following explosions at United States embassies in Kenya and Tanzania.

The bombings took place within minutes of each other at around 1030 local time.

The first blast happened in the Tanzanian capital Dar es Salaam and the second, just five minutes later, in Nairobi, Kenya’s capital city.

The Nairobi explosion demolished a five-story office block sending it crashing onto the embassy next door.

The US Ambassador Prudence Bushnell was meeting Kenyan Trade Minister Joseph Kamotho at the nearby Ufundi Cooperative Bank at the time but was only slightly injured.

The blast could be heard 10 miles (16km) away and caused total chaos in the city centre.

No-one has claimed responsibility but US officials suspect the attacks were the work of Osama bin Laden, an Islamic fundamentalist.


A cloud the size of a man’s fist on the horizon. Earlier in the year, India and then Pakistan had carried out a series of underground nuclear tests, which “provoked worldwide condemnation and fears of a nuclear conflict in one of the world’s most volatile regions.”

Which just goes to show how wrongly placed fears can be.
unique link to this extract

Former Coinbase official Balaji Srinivasan closes out $1m bitcoin bet early • Bloomberg via Yahoo

Olga Kharif:


Balaji Srinivasan, the former chief technology officer of Coinbase Global Inc., said he closed out what appeared to be a losing bet that Bitcoin would rise to $1m within 90 days.

Srinivasan said he gave $1m to two organizations, including the Bitcoin Core development team at researcher Chaincode Labs, as well as paying $500,000 to someone who goes by James Medlock on Twitter, and who won the wager.

The goal of the bet, Srinivasan reiterated in a Twitter post and a short video Tuesday, was to show that fiat currencies such as the dollar are in trouble, and that those troubles will push Bitcoin’s price up. At $28,710, Bitcoin is about 10% up from when Srinivasan accepted the bet on March 17.

The terms of the wager weren’t immediately clear. Medlock and Srinivasan didn’t return requests for comment.

“The reason that I did that is I wanted to tell you in a provable way that there’s something wrong in the economy and the state isn’t telling you about it,” Srinivasan said in the video, recounting troubles with US banks, sovereign debt and other potential issues. “That is what I am doing at my own expense, I am raising public alarm.”

Back on March 16, Medlock posted a tweet, “I’ll bet anyone $1m dollars that the US does not enter hyperinflation.”

On March 17, Srinivasan responded with, “I will take that bet.”


Honestly, how did someone with so little comprehension of the world get so rich that he can afford to throw a million dollars away?
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: 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 )

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.