Start Up No.1934: CNet’s AI bodysnatchers, Amazon frowns on Smile, Musk and the ‘self-driving’ Tesla, NFT auctions, and more

The UK loses a lot of electricity generated by wind farms in Scotland due to insufficient electricity infrastructure. What’s the solution? CC-licensed photo by Michael Coghlan 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.

There’s another post coming at the Social Warming Substack at about 0845 UK time, looking at the question of whether social media does influence elections. Or get it in your inbox! Free signup.

A selection of 10 links for you. Me, a bot? I’m @charlesarthur on Twitter. Observations and links welcome.

Inside CNET’s AI-powered SEO money machine • The Verge

Mia Sato and James Vincent:


Beyond the small CNET Money team, few people at the outlet know specific details about the AI tools — or the human workflow around them — that outraged readers last week, according to current and former staffers who spoke to The Verge on the condition that they remain anonymous. Under the two-year-old management of a private equity company called Red Ventures, CNET’s editorial staff has often been left wondering: was this story written by AI or a co-worker? Even today, they’re still not sure.

CNET was once a high-flying powerhouse of tech reporting that commanded a $1.8bn purchase price when it was acquired by CBS in 2008. Since then, it has fallen victim to the same disruptions and business model shifts as the rest of the media industry, resulting in CBS flipping the property to Red Ventures for just $500m in 2020.

Red Ventures’ business model is straightforward and explicit: it publishes content designed to rank highly in Google search for “high-intent” queries and then monetizes that traffic with lucrative affiliate links. Specifically, Red Ventures has found a major niche in credit cards and other finance products. In addition to CNET, Red Ventures owns The Points Guy, Bankrate, and, all of which monetize through credit card affiliate fees.

The CNET AI stories at the center of the controversy are straightforward examples of this strategy: “Can You Buy a Gift Card With a Credit Card?” and “What Is Zelle and How Does It Work?” are obviously designed to rank highly in searches for those topics. Like CNET, Bankrate and have also published AI-written articles about credit cards with ads for opening cards nestled within. Both Bankrate and directed questions about the use of AI to Lance Davis, the vice president of content at Red Ventures; CNET’s disclosure also included Davis as a point of contact until last week.

This type of SEO farming can be massively lucrative.


There’s a big story here, and Sato and Vincent poke at it effectively: the writers aren’t quite sure what stuff is generated by bots, but it’s a growing amount. Basically, a journalistic version of Invasion of the Body Snatchers.
unique link to this extract

Amazon’s Smile: gone after 10 years • Motley Fool via MSN

Anders Bylund on Amazon’s decision to shut down its charitable giving system:


let’s look at data from fiscal year 2020 — the latest period for which I found documentation of Amazon Smile’s actual donations. The parent company’s online retail operations generated $216bn of sales that year. A 0.5% cut of that massive revenue flow amounts to $1.08bn. But the AmazonSmile Foundation’s Form 990 tax filing showed just $73.6m in revenue that year.

In other words, roughly 0.03% of Amazon’s total sales passed through the Smile filter in 2020. So the Smile program sounds generous at first blush, thanks to Amazon’s ginormous scale, but the company never put its back into this effort. The resulting charitable contributions are reportedly hardly worth the paperwork and promotional work the charities have to put in.

All told, Amazon Smile has generated donations of $449m in 10 years. That’s commendable, but $40m a year is just a rounding error on Amazon’s income statement.

Critics of the program have lambasted Amazon for setting up a donation system that isn’t active by default, gives donation-based tax credits to Amazon instead of the shopper, and requires charities to promote Amazon’s shopping portal if they want people to send their Smile contributions to a specific cause.

Again, it’s not like a few million dollars of donation credits per year makes much of a difference to Amazon’s tax bills. In the example year of 2020, Amazon sent $1.71bn of cash to Uncle Sam and other income tax authorities. You could add back the Amazon Smile credits and you would barely notice the difference.


Once again, this shows the power of defaults. You had to go hunting for Smile, especially on the phone app. As the calculation shows, if Smile had been the default and you had to opt out, it would have generated a colossal amount of money.
unique link to this extract

Musk oversaw video that exaggerated Tesla’s self-driving capabilities • Bloomberg via Yahoo

Dana Hull and Sean O’Kane:


Elon Musk oversaw the creation of a 2016 video that exaggerated the abilities of Tesla Inc.’s driver-assistance system Autopilot, even dictating the opening text that claimed the company’s car drove itself, according to internal emails viewed by Bloomberg.

Musk wrote to Tesla’s Autopilot team after 2 a.m. California time in October 2016 to emphasize the importance of a demonstration drive to promote the system, which the chief executive officer made a splashy announcement about a week later. In an Oct. 19 call with reporters and blog post, Tesla said that all its cars from that day forward would ship with the hardware necessary for full self-driving capability.

“Just want to be absolutely clear that everyone’s top priority is achieving an amazing Autopilot demo drive,” Musk said in the email. “Since this is a demo, it is fine to hardcode some of it, since we will backfill with production code later in an OTA update,” he wrote, referring to using temporary code and updating it later using an over-the-air [OTA] software update.

“I will be telling the world that this is what the car *will* be able to do,” Musk continued, “not that it can do this upon receipt.”

The email sheds light on Musk’s mindset before he and Tesla then made claims about capabilities that have yet to materialize more than six years later. After cycling through several different iterations of hardware, the company to this day tells customers using Autopilot and the system it markets as Full Self-Driving to keep their hands on the wheel and be prepared to take over at any moment.

…Nine days later, after Tesla staffers shared a fourth version of the video, Musk replied that there were still too many jump cuts, and that the demo footage “needs to feel like one continuous take.”

While Musk had written in the earlier email that he would be clear Tesla was demonstrating what its cars would be able to do in the future, he then instructed staffers to open the video with a black screen and three sentences referring to the present.


unique link to this extract

What happens when AI has read everything? • The Atlantic

Ross Andersen:


Writing has allowed human beings to capture and store a great many more of our words. But like most new technologies, writing was expensive at first, which is why it was initially used primarily for accounting. It took time to bake and dampen clay for your stylus, to cut papyrus into strips fit to be latticed, to house and feed the monks who inked calligraphy onto vellum. These resource-intensive techniques could preserve only a small sampling of humanity’s cultural output.

Not until the printing press began machine-gunning books into the world did our collective textual memory achieve industrial scale. Researchers at Google Books estimate that since Gutenberg, humans have published more than 125 million titles, collecting laws, poems, myths, essays, histories, treatises, and novels. The Epoch team estimates that 10 million to 30 million of these books have already been digitized, giving AIs a reading feast of hundreds of billions of, if not more than a trillion, words.

Those numbers may sound impressive, but they’re within range of the 500 billion words that trained the model that powers ChatGPT. Its successor, GPT-4, might be trained on tens of trillions of words. Rumors suggest that when GPT-4 is released later this year, it will be able to generate a 60,000-word novel from a single prompt.

Ten trillion words is enough to encompass all of humanity’s digitized books, all of our digitized scientific papers, and much of the blogosphere. That’s not to say that GPT-4 will have read all of that material, only that doing so is well within its technical reach. You could imagine its AI successors absorbing our entire deep-time textual record across their first few months, and then topping up with a two-hour reading vacation each January, during which they could mainline every book and scientific paper published the previous year.


The thing about generating novels from a single prompt is concerning. You can imagine that Amazon’s Kindle Store is about to become overwhelmed. Well, even more overwhelmed.
unique link to this extract

Supreme Court poised to reconsider key tenets of online speech • The New York Times

David McCabe:


On Friday, the Supreme Court is expected to discuss whether to hear two cases that challenge laws in Texas and Florida barring online platforms from taking down certain political content. Next month, the court is scheduled to hear a case that questions Section 230, a 1996 statute that protects the platforms from liability for the content posted by their users.

The cases could eventually alter the hands-off legal position that the United States has largely taken toward online speech, potentially upending the businesses of TikTok, Twitter, Snap and Meta, which owns Facebook and Instagram.

“It’s a moment when everything might change,” said Daphne Keller, a former lawyer for Google who directs a program at Stanford University’s Cyber Policy Center.

The cases are part of a growing global battle over how to handle harmful speech online. In recent years, as Facebook and other sites attracted billions of users and became influential communications conduits, the power they wielded came under increasing scrutiny. Questions arose over how the social networks might have unduly affected elections, genocides, wars and political debates.

In some parts of the world, lawmakers have moved to rein in the platforms’ influence over speech. Last year, European legislators approved rules that require internet companies to carry out procedures for taking down illicit content and to be more transparent about how they recommend content to people.

In the United States, where freedom of speech is enshrined in the First Amendment, there has been less legislative action. While lawmakers in Washington have grilled the chief executives of the tech giants over the past three years about the content they take down, proposals to regulate harmful content haven’t gotten traction.

Partisanship has made the logjam worse. Republicans, some of whom have accused Facebook, Twitter and other sites of censoring them, have pressured the platforms to leave more content up. In contrast, Democrats have said the platforms should remove more content, like health misinformation.


Arguably it’s the partisanship that has been saving Section 230. Quite what the social media landscape will look like if it’s repealed is anyone’s guess. Much messier, for sure.
unique link to this extract

Why Mastodon feels a lot like early Twitter • Fast Company

Jared Newman:


Twitter had a lot of problems in its early years. The site was notoriously unreliable, to the point that major news events, such as Michael Jackson’s death in June 2009 and the World Cup 2010, would crash the site. Hashtags and retweets were initially bolted on by users and didn’t become official features until July and November 2009, respectively. Twitter didn’t even have its own iPhone app until April 2010, when it acquired the popular third-party client, Tweetie.

The parallels to Mastodon should be obvious by now. Mastodon servers—particularly the major ones—have struggled to meet demand during usage spikes, and the official Mastodon apps are worse than third-party alternatives. Key Mastodon concepts, such as servers, local timelines, and content warnings, also can seem alien in the same way hashtags and retweets used to be.

The site also has other challenges to address. The onboarding experience really needs improvement, and the network would benefit by leaning into local timelines as a standout feature. More importantly, its moderators must get better at welcoming Blacks and marginalized groups—and at recognizing threats against them by bad actors.

It’s natural, then, for some portion of new users to not immediately understand the value of “not Twitter” and go back to their old habits. But having spent an increasing amount of time on Mastodon over the past couple of months—mostly at the expense of Twitter—the energy around the platform is hard to ignore.


Certainly it feels to me as though a lot of air has gone out of Twitter quite quickly, just recently. The cutting-off of third-party clients, forcing the use of the (terrible) first-party app, contributes: the rough edges of using that make the desolation feel even more acute.
unique link to this extract

Ransomware revenue down as more victims refuse to pay • Chainalysis

Chainalysis Team:


2022 was an impactful year in the fight against ransomware. Ransomware attackers extorted at least $456.8m from victims in 2022, down from $765.6m the year before. 

As always, we have to caveat these findings by noting that the true totals are much higher, as there are cryptocurrency addresses controlled by ransomware attackers that have yet to be identified on the blockchain and incorporated into our data. When we published last year’s version of this report, for example, we had only identified $602m in ransomware payments in 2021. Still, the trend is clear: Ransomware payments are significantly down.

However, that doesn’t mean attacks are down, or at least not as much as the drastic dropoff in payments would suggest. Instead, we believe that much of the decline is due to victim organizations increasingly refusing to pay ransomware attackers. We’ll discuss this phenomenon more below, but first, let’s look more at general ransomware trends in 2022.

Despite the drop in revenue, the number of unique ransomware strains in operation reportedly exploded in 2022, with research from cybersecurity firm Fortinet stating that over 10,000 unique strains were active in the first half of 2022. On-chain data confirms that the number of active strains has grown significantly in recent years, but the vast majority of ransomware revenue goes to a small group of strains at any given time. We do, however, see turnover throughout the year among the top-grossing strains.

Likewise, ransomware lifespans continue to drop. In 2022, the average ransomware strain remained active for just 70 days, down from 153 in 2021 and 265 in 2020. As we’ll explore below, this activity is likely related to ransomware attackers’ efforts to obfuscate their activity, as many attackers are working with multiple strains.


This reminds me of the point in the early 2000s when companies started to get on top of computer viruses as a source of trouble: the growth in antivirus and threat protection systems meant that any infection was short-lived and quickly cleaned up, and it became harder for outbreaks to spread. The drop in revenue is significant after two years when it looked set for much bigger numbers. Defences are improving, refusal to pay is growing, and while – like viruses – this problem isn’t going away, it’s going to fade into the background as a cost of doing business.
unique link to this extract

The UK is wasting a lot of wind power • Archy de Berker

Archy de Berker and Peter Dudfield:


we often have to move electricity from North to South. The map below, from the National Grid, shows areas of surplus energy generation in blue and those with a deficit in red, and the resulting transfers needed to balance the grid.

Source: National Grid Future Energy Scenarios 2022 

This poses a problem, because moving electricity long distances is expensive. You need big cables, which are serious bits of kit – the last large one we put in cost £1.2 bn. At times, we just have more windpower than we have cables to transmit it. The particular hotspot for this problem is the B6 boundary: the bottleneck for electricity from Scotland to flow to England.

When we’re generating more windpower than we can transmit, the National Grid pays the windfarms to turn off, and pays a (typically gas powered) alternative generator, closer to the demand, to turn on. Consumers end up effectively paying three times for the power they’re getting: the original payment to the windfarm for the electricity, the payment to turn off, and then the payment to the alternative generator.

In the past, it has made financial sense to avoid the expense of building extra cables and instead pay a bit more to replace the lost wind power with gas generation in the South. However, with gas prices surging, this doesn’t look like such a good trade off for consumers, not to mention the planet.

We’ve built an interactive dashboard for exploring curtailment [when the amount of power taken from turbines is so large it has to be capped, and wind farms are paid not to feed the surplus into the grid, and gas plants are turned on instead] in 2022. You can explore it here.

At times the UK was wasting as much wind power as we were using.

On Christmas day, we spent £9.2m on curtailment costs, curtailing a total of 76.18 GWh. That’s enough electricity to power ~11,000 households for a year.

…However, laying high voltage cables is slow – much slower than building new wind turbines. In the time it takes for this transmission to come online (2GW by 2027 & 4GW by 2029), we will have added far more new wind capacity North of the B6 boundary.

Simply put: we can’t lay cables fast enough to solve this problem.


Tricky problem. Think: how would you solve it? Then read the suggested solution.
unique link to this extract

Grenfell showed exactly why we need red tape, yet the Tories are desperate to bin thousands of laws • The Guardian

Peter Apps is the author of Show Me the Bodies: How We Let Grenfell Happen:


In the early 2000s, the UK government was warned that the country risked becoming the continent’s “dumping ground” for dangerous cladding and insulation products. At the time, civil servants were working on plans to harmonise fire standards with the rest of the European Union – but this harmonisation never happened.

Corporate lobbyists had argued against it, claiming there would be “economic consequences” if their members were unable to sell their combustible building products freely. As standards were tightened in much of Europe, the UK would not update its outdated guidance for the next 17 years.

The result was gaps in the regulation, which the free market was more than efficient enough to find and exploit. “The evolution of fire regulation will put [highly combustible cladding] out of the market in the coming month[s],” wrote a senior figure at cladding firm Arconic in 2009. “For the moment, even if we know that [the material] has a bad behaviour exposed to fire, we can still work with national regulations who are not as restrictive.”

One such market was the UK, which Arconic targeted with its violently combustible panels, knowing that their marginally lower cost would make them attractive. The panels were fitted on hundreds of tall buildings around the country.

One was Grenfell Tower. And in June 2017, it caught fire, the block was engulfed in flames and 72 people died horrible, avoidable deaths. It is important to recount this story now because the lessons have been ignored.

The government is now preparing to sweep away thousands of EU rules of the statue book in a single stroke. This is what prompted the former business secretary, Jacob Rees-Mogg, to gleefully tweet his approval this month of “igniting the deregulatory bonfire”.


James Dyson was similarly fulminating in a column for the Daily Telegraph about being stifled by regulation. Amazing how regulation is only bad when it affects rich people. I wrote about the necessity for regulation back in 2017, in relation to deaths at sea, and it’s still true.
unique link to this extract

SBF, Bored Ape Yacht Club, and the spectacular hangover after the art world’s NFT gold rush • Vanity Fair

Nate Freeman:


[The analyst Benedict] Evans offered another conundrum. When a market offers something for sale at a large sum, there is, at a base level, an understanding among the public that it has some legitimate importance. Perhaps the artwork is not to one’s liking, but it has a provenance and the artist is in museum collections—or there’s historic relevance to something that makes it at the very least a curio.

“When you are buying vintage vinyl, or rare sneakers, or Marilyn Monroe’s shoes, or a Salvador Dalí print, or whatever it is, you’re getting something that has no tangible physical value, but cultural value,” Evans said. “There’s like a deep cultural base that thinks Jordan sneakers are worth something, early Sex Pistols vinyl is worth something. And the challenge with all of these NFTs was you didn’t really know that there was that broad, deep cultural base. It was just: ‘Oh, my gosh, somebody just bought one for 200 grand.’”

For the time being, some in the art world are still acting as though the devotion to NFTs could result in some kind of windfall. Sotheby’s Metaverse has a sale up right now. It’s offering the first NFTs by the artist Sebastião Salgado, but they aren’t exactly lighting the place on fire. They cost $250 each. Back in 2021, the Natively Digital sale netted Sotheby’s $17m, with $11m paid for a single NFT from the series of CryptoPunks. 

But in February 2022, Sotheby’s set up a special sale where it expected a set of 104 CryptoPunks to go for as much as $30m, only to see the thing collapse minutes before the gavel-in when the consignor backed out, reportedly due to a lack of interest from bidders. By last December, the Natively Digital sale seemed to have lost its luster entirely. Sotheby’s offered the first-ever Keith Haring NFT as the star lot of the sale, but it sold for $25,000, well below the $80,000 high estimate.

…When asked for comment on whether they plan to continue with their NFT platforms, Sotheby’s declined. Christie’s did not respond to a request for comment.


A market down by 96% year-on-year. I’d like to say there’s no way it’s coming back, but rather as with foolproof systems that prove not to be, the art and crypto spaces keep coming up with fools who exceed our expectations.

It would have been good if Freeman could have cornered some of the auction people who rode this unicorn. But they’re too busy counting the money, one presumes. Or crying over their crypto losses, as quite a few of the transactions were made with those magic beans.
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.