Start Up No.1914: how generative AI will change work, 52 things you (probably) didn’t know, Tether’s loan trouble, and more

The Maersk shipping company is giving up a blockchain project with IBM because it lacks commercial viability. Are there any out there left? CC-licensed photo by Mohammad Rizky 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.

It’s Friday, so there’s another post due at the Social Warming Substack at about 0845 UK time.

A selection of 9 links for you. Not autocompleted. I’m @charlesarthur on Twitter. Observations and links welcome.

Generative AI: autocomplete for everything • Noahpinion

Noah Smith and “roon”, who works at one of the generative AI companies:


as roon likes to say, every time you use any of the most advanced AI applications, you’re “lighting a pile of GPUs on fire”. Those resource constraints explain why humans who want jobs will find jobs: AI businesses will just keep expanding and gobbling up more physical resources until human workers themselves, and the work they do to complement AI, become the scarce resource.

The principle of comparative advantage says that whether the jobs of the future pay better or worse than the jobs of today depends to some degree on whether AI’s skill set is very similar to humans, or complementary and different. If AI simply does things differently than humans do, then the complementarity will make humans more valuable and will raise wages. 

And although we can’t speak to the AI of the future, we believe that the current wave of generative AI does things very differently from humans. AI art tends to differ from human-made art in subtle ways – its minor details are often off in a compounding uncanny valley fashion that the net result can end up looking horrifying. Anyone who’s ridden in a Tesla knows that an AI backs into a parallel parking space differently than a human would. And for all the hype regarding large language models passing various forms of the Turing Test, it’s clear that their skillset is not exactly the same as a human’s.

Because of these differences, we think that the work that generative AI does will basically be “autocomplete for everything”. 

…Industrial design will work in a similar way. Take a look at any mundane, boring object in the room around you – a lamp, or a TV stand, or a coffee maker. Some human being had to come up with the design for that. With generative AI, the designer won’t have to look through pages and pages of examples to riff off of. They’ll just deliver a prompt – “55-inch TV stand with two cabinets” – and see a menu of alternative designs. They’ll pick one of the designs, refine it, and add any other touches they want.

We can imagine a lot of jobs whose workflows will follow a similar pattern – architecture, graphic design, or interior design. Lawyers will probably write legal briefs this way, and administrative assistants will use this technique to draft memos and emails. Marketers will have an idea for a campaign, generate copy en masse and provide finishing touches. Consultants will generate whole powerpoint decks with coherent narratives based on a short vision and then provide the details. Financial analysts will ask for a type of financial model and have an Excel template with data sources autofilled.

What’s common to all of these visions is something we call the “sandwich” workflow. This is a three-step process. First, a human has a creative impulse, and gives the AI a prompt. The AI then generates a menu of options. The human then chooses an option, edits it, and adds any touches they like.


unique link to this extract

IBM and Maersk abandon ship on TradeLens logistics blockchain • Coindesk

Danny Nelson:


Maersk and IBM will wind down their shipping blockchain TradeLens by early 2023, ending the pair’s five-year project to improve global trade by connecting supply chains on a permissioned blockchain.

TradeLens emerged during the “enterprise blockchain” era of 2018 as a high-flying effort to make inter-corporate trade more efficient. Open to shipping and freight operators, its members could validate the transaction of goods as recorded on a transparent digital ledger.

The idea was to save its member-shipping companies money by connecting their world. But the network was only as strong as its participants; despite some early wins, TradeLens ultimately failed to catch on with a critical mass of its target industry.

“TradeLens has not reached the level of commercial viability necessary to continue work and meet the financial expectations as an independent business,” Maersk Head of Business Platforms Rotem Hershko said in a statement.


Have to wonder what level of commercial viability it did actually reach. Also, hope someone out there is keeping tabs on all the blockchain projects that were announced since, oh, 2015, and how they’re faring. I seem to recall IBM being enormously keen on it at one point. Going to guess meanwhile that Maersk manages to keep tabs just fine on what’s on its ships.
unique link to this extract

52 things I learned in 2022 • Magnetic Notes

Tom Whitwell:


1. A bolt of lightning contains about ¼ of a kilowatt-hour of power. Even with recent energy price rises, it’s only worth about 9 pence. [Sarah Jensen]

2. A ‘zhènlóuqì’ is an electrical floor shaker sold on Taobao, used to get revenge on noisy neighbours. [Wang Xinyi]

3. In the UK and Australia, people tend to turn left when entering a building. In the US, they turn right. It’s important to remember if you’re booking a trade show booth. [Marc Abrahams]

4. Using ellipsis in writing signifes the writer is Gen-X or Boomer and can read as confusing, passive-aggressive or even weirdly flirtatious to digital natives. [Kaye Whitehead, from Gretchen McCulloch]

5. CountThings is an very successful app that counts things. It costs $120/month. The templates page shows the things people pay to count. [CountThings]

6. Heavenbanning is a hypothetical way to moderate social networks. Instead of being thrown off the platform, bad actors have all their followers replaced with sycophantic AI models that constantly agree and praise them. Real humans never interact with them. [Asara Near]


Plenty more where those came from, all fascinating. I think No.7 might be one of the best. Though also No.11. Wait, No.13. Oh..
unique link to this extract

Elon Musk’s Twitter polls are bot-driven B.S., ex-employees say • Rolling Stone

Noah Shachtman and Adam Rawnsley:


“One of the first products I worked on was polls. And one of the big discussions was around the tradeoffs between integrity and privacy – keeping logs [or each user’s vote] or not. We landed on the side of privacy,” Yoel Roth, Twitter’s former Head of Trust and Safety who resigned this month, told Rolling Stone. 

“Polls are more prone to manipulation than almost anything else [on Twitter]. It’s interesting, given his [Elon’s] use of polls,” he added. Several other ex-Twitter employees gave similar assessments.

Twitter did not immediately respond to questions from Rolling Stone, likely because Musk fired the company’s communications team.  

The reliance on bot-heavy polls is doubly ironic, given that Musk once balked at buying the company over concerns that there were too many inauthentic accounts on the platform. Now he’s all-but-counting on them in order to justify big decisions about Twitter’s future. 

“A Twitter poll can be manipulated. There’s nothing scientific or rigorous in any way about what he’s doing,” Sarah T. Roberts, a former Twitter employee and current faculty director for UCLA’s Center for Critical Internet Inquiry, told the Washington Post.

…When Musk acted as a self-appointed envoy between Russia and Ukraine, the billionaire displayed at least a dim understanding that the polling feature could theoretically be gamed when users voted down his proposal for a peace deal. “The bot attack on this poll is strong!” he replied to one fan as Twitter users panned his idea.


unique link to this extract

Biotech labs are using AI inspired by DALL-E to invent new drugs • MIT Technology Review

Will Douglas Heaven:


The explosion in text-to-image AI models like OpenAI’s DALL-E 2—programs trained to generate pictures of almost anything you ask for—has sent ripples through the creative industries, from fashion to filmmaking, by providing weird and wonderful images on demand.

The same technology behind these programs is also making a splash in biotech labs, which are increasingly using this type of generative AI, known as a diffusion model, to conjure up designs for new types of protein never seen in nature.

On Thursday, two labs separately announced programs that use diffusion models to generate designs for novel proteins with more precision than ever before. Generate Biomedicines, a Boston-based startup, revealed a program called Chroma, which the company describes as the “DALL-E 2 of biology.”

At the same time, a team at the University of Washington led by biologist David Baker has built a similar program called RoseTTAFold Diffusion. In a preprint paper posted online today, Baker and his colleagues show that their model can generate precise designs for novel proteins that can then be brought to life in the lab. “We’re generating proteins with really no similarity to existing ones,” says Brian Trippe, one of the co-developers of RoseTTAFold.

These protein generators can be directed to produce designs for proteins with specific properties, such as shape or size or function. In effect, this makes it possible to come up with new proteins to do particular jobs on demand. Researchers hope that this will eventually lead to the development of new and more effective drugs. “We can discover in minutes what took evolution millions of years,” says Gevorg Grigoryan, CEO of Generate Biomedicines.

…Generating strange designs on a computer is one thing. But the goal is to turn these designs into real proteins. To test whether Chroma produced designs that could be made, Generate Biomedicines took the sequences for some of its designs—the amino acid strings that make up the protein—and ran them through another AI program. They found that 55% of them would be predicted to fold into the structure generated by Chroma, which suggests that these are designs for viable proteins.

Baker’s team ran a similar test. But Baker and his colleagues have gone a lot further than Generate Biomedicines in evaluating their model. They have created some of RoseTTAFold Diffusion’s designs in their lab.


I think it’s pushing it to say the systems are “inspired by DALL-E”. As he writes, it’s the same technology – generative adversarial networks – but starting from very different points, with totally different training data. Sure, Chroma is trying to get attention by using that line, but don’t be fooled.

unique link to this extract

⚠️ Warning: do not use Hive Social 👉🐝👈 • zerforschung


Following the Twitter takeover, a number of services promising to be an alternative gained traction. One of those is “Hive Social”, which reached more than a million users in the last weeks.

Of course, we were interested and took a look at Hive from a security standpoint. We found a number of critical vulnerabilities, which we confidentially reported to the company. After multiple attempts to contact the company we finally reached them by phone and they acknowledged the report. After multiple days and multiple reminders by us, they claimed to fix them within the next two days. However after those two days, multiple vulnerabilities we reported were not fixed and still existed at the time of writing.

⚠️ We strongly advise against using Hive in any form in the current state.

The issues we reported allow any attacker to access all data, including private posts, private messages, shared media and even deleted direct messages. This also includes private email addresses and phone numbers entered during login.


Then there’s an update: “The vulnerabilities are currently no longer exploitable because Hive deactivated their servers.”

A bit unsurprising that an app which claims to have only two, possibly three, people writing it should turn out to have a huge security hole in it.
unique link to this extract

Mark Zuckerberg says Apple’s policies are not “sustainable” • Axios

Sara Fischer:


Zuckerberg has been one of the loudest critics of Apple in Silicon Valley for the past two years. In the wake of Elon Musk’s attacks on Apple this week, his concerns are being echoed more broadly by other industry leaders and Republican lawmakers.

“I think the problem is that you get into it with the platform control, is that Apple obviously has their own interests,” Zuckerberg said at The New York Times’ Dealbook conference.

“[T]he fact that companies have to deliver their apps exclusively through platforms that are controlled by competitors — there is a conflict of interest there,” he said. That conflict of interest makes Apple “not just a kind of governor that is looking out for the best of people’s interests.”

Zuckerberg also noted that Apple’s policies differ from other tech giants, including Microsoft and Google, which allow apps to be sideloaded onto devices if they’re inaccessible in app stores.

“I do think Apple has sort of singled themselves out as the only company that is trying to control, unilaterally, what apps get on the device and I don’t think that’s a sustainable or a good place to be.”

Changes to Apple’s app tracking policies last year are expected to cost Meta billions of dollars in lost ad revenue.


So if we really boil it down to what’s practicable, he’s saying Apple should allow sideloading. But that wouldn’t prevent Apple implementing Ad Tracing Transparency (ATT), which is his real bugbear. ATT works at the app level as well as on Safari.
unique link to this extract

Rising Tether loans add risk to stablecoin, crypto world • WSJ

Jonathan Weil:


The company behind the tether stablecoin has increasingly been lending its own coins to customers rather than selling them for hard currency upfront. The shift adds to risks that the company may not have enough liquid assets to pay redemptions in a crisis.

Tether Holdings Ltd. says it lends only to eligible customers and requires that borrowers post lots of “extremely liquid” collateral, which could be sold for dollars if borrowers default.

These loans have appeared for several quarters in the financial reports that Tether shows on its website. In the most recent report, they reached $6.1bn as of Sept. 30, or 9% of the company’s total assets. They were $4.1bn, or 5% of total assets, at the end of 2021. 

Tether calls them “secured loans” and discloses little about the borrowers or the collateral accepted. Alex Welch, a Tether spokeswoman, confirmed that all of the secured loans listed in the reports were issued and denominated in tether.

…The premise of tether—and other stablecoins—is that the issuer always will redeem one coin for $1. Issuers take pains to demonstrate they have ample funds available to do so.

The company’s reports show only US dollar amounts for the loans and don’t say the loans were made in tether tokens. The reports also say the loans were “fully collateralized by liquid assets.”

“I’ve been very skeptical and in disbelief that they can get away with the lack of disclosure and with the limited transparency,” said Peter Crane, president of Crane Data, which tracks money-market funds. “If you do have reserves, why wouldn’t you show them?” Both money-market funds and stablecoins like tether are supposed to maintain a value of $1.

The vast majority of the assets listed in Tether’s reports are in cash, Treasury bonds and other safe instruments easily converted to dollars. Loans are different. Tether can’t be certain the loans will be paid back, that it could sell the loans to a buyer for dollars in a pinch or that the collateral it holds will be adequate.


Tether’s PR person is extremely reassuring, but at this point I wouldn’t trust her to know what the truth is. For a long time in the crypto world Tether has been the dog that didn’t bark. Perhaps now, to mix a metaphor, it’s clearing its throat.
unique link to this extract

World of Moose ‘”2022 Advent Calendar” in a Robot Voice’ • worldofmoose


It’s a gift from Moose & Karen, it’s absolutely FREE of charge, all we ask is that you:

• have fun with it and do your very best colouring 

• share the link with your friends and get them to do it too 

• don’t sell it, alter it or use any of the images for anything else.


If you don’t follow @MooseAllain (that’s his name!) on Twitter, you really should. He’s very funny/punny (“love of an anagram is the one thing that unties us all”) and his cartoons are terrific too. At a time when advent calendars are either madly expensive or impossible to find, this one at least is satisfying for children and adults, and comparatively cheap.
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 )

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.