Start Up No.1984: FBI arrests accused leaker, music labels warn on AI-generated songs, tracing bitcoin thieves, stock Twitter?, and more

A coroner has warned that millions of people are at risk of death if they use full-face snorkelling masks. CC-licensed photo by Enrico Strocchi 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. Try another. I’m @charlesarthur on Twitter. On Mastodon: Observations and links welcome.

Air National Guardsman arrested as FBI searches his home • The New York Times

Aric Toler, Michael Schwirtz, Haley Willis, Riley Mellen, Christiaan Triebert, Malachy Browne, Thomas Gibbons-Neff and Julian E. Barnes:


Although gaming friends would not identify the leader of an online group linked to the leak of classified United States intelligence files, a trail of digital evidence compiled by The New York Times led toward Jack Teixeira, a 21-year-old member of the intelligence wing of the Massachusetts Air National Guard.

An online gaming profile in Airman Teixeira’s name connected him to photographs of the location where leaked documents were photographed — a kitchen countertop inside his childhood home.

An image shows a leaked document on a background that appears to be a textured countertop above a white tile floor.

The granite countertop bearing a distinctive pattern and white floor tiles are visible in the margins of some of the leaked intelligence briefings. The same interior décor is visible in photographs of the family home posted online by one of Airman Teixeira’s immediate relatives.

Mr. Teixeira was arrested by the FBI in North Dighton, Massachusetts, on Thursday afternoon.


The main link is to the full-length story with everything about yesterday’s arrest, but this is the part that is most relevant. On Tuesday, bellingcat said that “Creases can be seen on the documents with items, such as a hunter’s scope box and some Gorilla Glue visible in the background of those dated from early March”. To which I commented that “With details like that, bellingcat will probably pin down precisely when and where they were photographed in a few days.”

The top byline on Thursday’s story above: Aric Toler of bellingcat.
unique link to this extract

Streaming services urged to clamp down on AI-generated music • Financial Times

Anna Nicolaou:


Universal Music Group has told streaming platforms, including Spotify and Apple, to block artificial intelligence services from scraping melodies and lyrics from their copyrighted songs, according to emails viewed by the Financial Times.

UMG, which controls about a third of the global music market, has become increasingly concerned about AI bots using their songs to train themselves to churn out music that sounds like popular artists.

AI-generated songs have been popping up on streaming services and UMG has been sending takedown requests “left and right”, said a person familiar with the matter. The company is asking streaming companies to cut off access to their music catalogue for developers using it to train AI technology.

“We will not hesitate to take steps to protect our rights and those of our artists,” UMG wrote to online platforms in March, in emails viewed by the FT.

“This next generation of technology poses significant issues,” said a person close to the situation. “Much of [generative AI] is trained on popular music. You could say: compose a song that has the lyrics to be like Taylor Swift, but the vocals to be in the style of Bruno Mars, but I want the theme to be more Harry Styles. The output you get is due to the fact the AI has been trained on those artists’ intellectual property.” 


Much like the claim that Getty is bringing against Stable Diffusion on roughly this basis: it was trained on our content, so it’s breaking our copyright. (I’m not convinced Getty will prevail, but there’s a long way to go yet.)
unique link to this extract

Google will shut down Currents, the work-focused Google Plus replacement • The Verge

Mitchell Clark:


Google has announced that it’ll shut down Currents, which was introduced in 2019 as a replacement for Google Plus for G Suite. In a blog post, the company says it’s “planning to wind down” Currents, and that it’ll push the people who were using it to Spaces, which is sort of like Google Chat’s version of a Slack channel or Discord room.

Google says that it’s making the change so users won’t have to work in a “separate, siloed destination” — instead, they’ll be using Chat and Spaces, which will soon be prominently integrated into Gmail. The company promises it’s going to make Spaces a more suitable replacement over the next year, saying it’ll “deliver new capabilities” like “support for larger communities and leadership communication, investments in advanced search, tools for content moderation, and more.”

…Currents hasn’t gotten a lot of love from Google. I was only able to find three blog posts about it on Google’s Workspace Updates site — the one announcing it, one in 2020 announcing that it was widely available, and the one from Thursday announcing it was being shut down. It was included on Google’s main list of apps that come with Workspace at one point according to the WayBack Machine, but it seems like it was removed sometime in November 2021.


Another one bites the dust. Along with some of the job cuts, maybe Google could figure out how to focus on products and iteratively improve them, rather than binning them and expecting people to migrate. Microsoft has managed it with Teams, for example.
unique link to this extract

The US cracked a $3.4bn crypto heist—and bitcoin’s anonymity • WSJ

Robert McMillan:


[James] Zhong moved the stolen bitcoins [accrued via a bug in the Silk Road dark web site] from one account to another for eight years to cover his tracks. By late 2021, the red-hot crypto market had raised the value of his trove to $3.4bn [from $600,000 when he exploited it in 2012]. He still lived in a modest house in Athens, Ga., and dressed in shorts and T-shirts. He also had a lake-house getaway in Gainesville, Ga., a Lamborghini sports car and a $150,000 Tesla.

In November 2021, federal agents surprised Mr. Zhong with a search warrant and found the digital keys to his crypto fortune hidden in a basement floor safe and a popcorn tin in the bathroom. Mr. Zhong, who pleaded guilty to wire fraud, is scheduled to be sentenced Friday in New York federal court, where prosecutors are seeking a prison sentence of less than two years.

Mr. Zhong’s case is one of the highest-profile examples of how federal authorities have pierced the veil of blockchain transactions. Private and government investigators can now identify wallet addresses associated with terrorists, drug traffickers, money launderers and cybercriminals, all of which were supposed to be anonymous. 

Law-enforcement agencies, working with cryptocurrency exchanges and blockchain-analytics companies, have compiled data gleaned from earlier investigations, including the Silk Road case, to map the flow of cryptocurrency transactions across criminal networks worldwide. In the past two years, the US has seized more than $10bn worth of digital currency through successful prosecutions, according to the Internal Revenue Service—in essence, by following the money. Instead of subpoenas to banks or other financial institutions, investigators can look to the blockchain for an instant snapshot of the money trail. 


So bitcoin turns out to be the extremely nonymous currency, because it hasn’t expanded into real-world uses, which means that any time someone tries to cash out to useful currency, the exchanges are obliged to keep details of who’s doing the transaction. And they’ll get people with badges knocking at the door. (Link should work for free.)
unique link to this extract

Arrest made in SF killing of Bob Lee: tech exec’s alleged killer also worked in tech • Mission Local

Joe Eskenazi:


Mission Local is informed that the San Francisco Police Department early this morning made an arrest in the April 4 killing of tech executive Bob Lee, following an operation undertaken outside the city’s borders. The alleged killer also works in tech and is a man Lee purportedly knew.

We are told that police today were dispatched to Emeryville with a warrant to arrest a man named Nima Momeni. The name and Emeryville address SFPD officers traveled to correspond with this man, the owner of a company called Expand IT.

Multiple police sources have described the predawn knifing last week, which left the 43-year-old Lee dead in a deserted section of downtown San Francisco, as neither a robbery attempt nor a random attack. 

Rather, Lee and Momeni were portrayed by police as being familiar with one another. In the wee hours of April 4, they were purportedly driving together through downtown San Francisco in a car registered to the suspect. 

Some manner of confrontation allegedly commenced while both men were in the vehicle, and potentially continued after Lee exited the car. Police allege that Momeni stabbed Lee multiple times with a knife that was recovered not far from the spot on the 300 block of Main Street to which officers initially responded. 


Lee’s killing stirred up lots of strong feelings among the tech community: he was clearly widely adored by many who had encountered him. Yet there were details about the stabbing that were out of kilter: why was he out on his own at 3am, in a relatively safe part of the city, and had he really been stabbed randomly, given that nothing had been stolen from him?

The usual assumption about murder – you were killed by someone you know – seems to have prevailed.
unique link to this extract

Twitter to let users access stocks, crypto via eToro in finance push • CNBC

Ryan Browne:


Twitter will let its users access stocks, cryptocurrencies and other financial assets through a partnership with eToro, a social trading company.

Starting Thursday, a new feature will be rolled out on the Twitter app. It will allow users to view market charts on an expanded range of financial instruments and buy and sell stocks and other assets from eToro, the company told CNBC exclusively.

Currently, it’s already possible to view real-time trading data from TradingView on index funds like the S&P 500 and shares of some companies such as Tesla. That can be done using Twitter’s “cashtags” feature — you search for a ticker symbol and insert dollar sign in front of it, after which the app will show you price information from TradingView using an API (application programming interface).

With the eToro partnership, Twitter cashtags will be expanded to cover far more instruments and asset classes, an eToro spokesperson told CNBC.

You’ll also be able to click a button that says “view on eToro,” which takes you through to eToro’s site, and then buy and sell assets on its platform. EToro uses TradingView as its market data partner.


Chasing the stonk (sic) market. This will wipe out the crypto scams, right?
unique link to this extract

Elon Musk’s free-speech charade is over • The Atlantic

Adam Serwer:


until Musk bought Twitter late last year, conservatives were arguing that the company’s moderation decisions violated the First Amendment, even though Twitter is a private company and not part of the government. Now that Musk is using his editorial discretion as owner of the company to promote people and ideas he supports—primarily right-wing influencers—and diminish the reach of those he does not, the constitutional emergency has subsided. At least until his allies and defenders on Substack found themselves unable to promote their work on Twitter, free speech had been restored, because “free speech” here simply means that right-wing ideas and arguments are favoured. This outcome—that Twitter under Musk would favour right-wing content—was predictable, and I’m saying that because I wrote last April that that’s what would happen.

The episode reveals something important about the way that many conservative jurists and legal scholars now approach the principle of free speech. Florida and Texas passed laws prohibiting social-media companies from moderating user-generated content, in retaliation for what they characterized as liberal “censorship.” A federal judge appointed by Trump, Andrew Oldham, then upheld the Texas law with a ruling that scoffed at the idea that “editorial discretion” constituted a “freestanding category of First-Amendment-protected expression” and insisted that the platforms’ moderation decisions did not qualify for that protection.


Hypocrisy on a gigantic scale. It’s what Musk does best.
unique link to this extract

The mounting human and environmental costs of generative AI • Ars Technica

Sasha Luccioni is a researcher and climate lead at Hugging Face:


For large language models (LLMs) like ChatGPT, we’ve gone from around 100 million parameters in 2018 to 500 billion in 2023 with Google’s PaLM model. The theory behind this growth is that models with more parameters should have better performance, even on tasks they were not initially trained on, although this hypothesis remains unproven.

Bigger models typically take longer to train, which means they also need more GPUs, which cost more money, so only a select few organizations are able to train them. Estimates put the training cost of GPT-3, which has 175 billion parameters, at $4.6m—out of reach for the majority of companies and organisations. (It’s worth noting that the cost of training models is dropping in some cases, such as in the case of LLaMA, the recent model trained by Meta.)

This creates a digital divide in the AI community between those who can train the most cutting-edge LLMs (mostly Big Tech companies and rich institutions in the Global North) and those who can’t (nonprofit organisations, startups, and anyone without access to a supercomputer or millions in cloud credits).

Building and deploying these behemoths requires a lot of planetary resources: rare metals for manufacturing GPUs, water to cool huge data centers, energy to keep those data centres running 24/7 on a planetary scale… all of these are often overlooked in favour of focusing on the future potential of the resulting models.


$4.6m doesn’t sound wildly beyond the funding possibilities of some Silicon Valley startups. But increasingly I feel like we’re heading towards IBM president Thomas Watson’s prediction from 1943 – “I think there is a world market for maybe five computers.” Oh, but what computers.
unique link to this extract

Full-face snorkel caused death of Red Sea tourist • The Times

Jonathan Ames:


Full-face snorkel masks could be putting millions of people at risk, a coroner has warned after the death of a British tourist in Egypt.

Angela Kearn, an experienced snorkeller, collapsed and died at a Red Sea resort while using equipment from the retailer Decathlon, an inquest was told. A coroner found that she had developed immersion pulmonary oedema — a concentration of fluid in her lungs — and that the mask “contributed” to her difficulties.

Kearn, 63, had recently been diagnosed with high blood pressure and was taking hormone replacement therapy to relieve symptoms of menopause, the inquest was told. She was unaware that this made her vulnerable when using the mask, a point that was allegedly not made clear in the packaging.

Caroline Topping, the assistant coroner for Surrey, has written to the General Medical Council, National Trading Standards and Decathlon in the UK to raise concerns about the potential threat the masks pose. An estimated 16 million people have bought them.

In a prevention of future deaths report, Topping wrote: “Many millions of the full-face masks have been sold and the safety concerns about their use by those with ongoing cardiovascular and respiratory issues has not been widely publicised or brought to the attention of those who already own the masks.”


Sixteen million. Even if the death rate is only one in a million, that’s 16 avoidable deaths. To make matters worse, there’s no British Standard for these masks – so you don’t know how safe or dangerous they might be for you.
unique link to this extract

Teens don’t care as much about VR as tech companies do • Fortune

Prarthana Prakash:


Digitally obsessed teens are a target market for companies selling virtual reality headsets that transport users into entirely new worlds. But it turns out, the younger set isn’t particularly interested—even though tech companies are spending billions on VR. 

Very few teens use VR devices daily, according to a study published Tuesday by investment bank Piper Sandler. The bank surveyed over 5,600 US teens and found that close to a third of them owned VR devices, but only 4% use them every day and only 14% use them every week. 

“To us, the lukewarm usage demonstrates that VR remains ‘early days’ and that these devices are less important than smartphones,” Piper Sandler analysts said in their report.

The report noted that between the second half of last year and now, use of VR devices has remained flat. Compared to the same time last year, use has actually dropped to 14% compared to 17%. 

Piper Sandler’s results reflect the relatively slow adoption of VR among younger people given that 95% of teens in the US own smartphones and 80% own gaming consoles, according to Pew Research.


OK, you could have said this about smartphone ahead of the iPhone’s launch, and about tablets ahead of the iPad’s launch, and about wireless headphones ahead of the Airpods’ launch. Even so, this doesn’t feel like a market waiting to happen.
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.