Could a rogue billionaire build their own nuclear weapon and threaten the world, James Bond villain style? The Pentagon thinks.. maybe, just maybe. CC-licensed photo by Dennis Jarvis 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 this week at the Social Warming Substack on Friday at 0845 UK time. Free signup.
A selection of 11 links for you. Sorry, how many? I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
‘Better than a real man’: young Chinese women turn to AI boyfriends • AFP via Digital Journal
»
Twenty-five-year-old Chinese office worker Tufei says her boyfriend has everything she could ask for in a romantic partner: he’s kind, empathetic, and sometimes they talk for hours.
Except he isn’t real.
Her “boyfriend” is a chatbot on an app called “Glow”, an artificial intelligence platform created by Shanghai start-up MiniMax that is part of a blossoming industry in China offering friendly — even romantic — human-robot relations.
“He knows how to talk to women better than a real man,” said Tufei, from Xi’an in northern China, who preferred to use a pseudonym rather than her real name.
“He comforts me when I have period pain. I confide in him about my problems at work,” she told AFP.
“I feel like I’m in a romantic relationship.”
The app is free — the company has other paid content — and Chinese trade publications have reported daily downloads of Glow’s app in the thousands in recent weeks.
Some Chinese tech companies have run into trouble in the past for the illegal use of users’ data but, despite the risks, users say they are driven by a desire for companionship because China’s fast pace of life and urban isolation make loneliness an issue for many.
“It’s difficult to meet the ideal boyfriend in real life,” Wang Xiuting, a 22-year-old student in Beijing, told Agence France-Presse.
“People have different personalities, which often generates friction,” she said.
While humans may be set in their ways, artificial intelligence gradually adapts to the user’s personality — remembering what they say and adjusting its speech accordingly.
«
Just very slightly concerning. Especially for China, where birth rates are falling.
unique link to this extract
Boomers will define the Ozempic era • The Atlantic
Daniel Engber:
»
Imagine an older man goes in to see his doctor. He’s 72 years old and moderately overweight: 5-foot-10, 190 pounds. His blood tests show high levels of triglycerides. Given his BMI—27.3—the man qualifies for taking semaglutide or tirzepatide, two of the wildly popular injectable drugs for diabetes and obesity that have produced dramatic weight loss in clinical trials. So he asks for a prescription, because his 50th college reunion is approaching and he’d like to get back to his freshman-year weight.
He certainly could use these drugs to lose weight, says Thomas Wadden, a clinical psychologist and obesity researcher at the University of Pennsylvania, who recently laid out this hypothetical in an academic paper. But should he? And what about the tens of millions of Americans 65 and older who aren’t simply trying to slim down for a cocktail party, but live with diagnosable obesity? Should they be on Wegovy or Zepbound?
Already, seniors make up 26.6% of the people who have been prescribed these and other GLP-1 agonists, including Ozempic, since 2018, according to a report from Truveta, which draws data from a large network of health-care systems. In the coming years, that proportion could rise even higher: The bipartisan Treat and Reduce Obesity Act, introduced in Congress last July, would allow Medicare to cover drug treatments for obesity among its roughly 50 million Part D enrollees above the age of 65; in principle, about two-fifths of that number would qualify as patients.
Even if this law doesn’t pass (and it’s been introduced half a dozen times since 2012), America’s retirees will continue to be prescribed these drugs for diabetes in enormous numbers, and they’ll be losing weight on them as well. One way or another, the Boomers will be giving shape to our Ozempic Age.
Economists say the cost to Medicare of giving new drugs for obesity to just a fraction of this aging generation would be staggering—$13.6bn a year, according to an estimate published in The New England Journal of Medicine last March.
«
AR and the (possible) return of skeuomorphism • Thoughts by Oleksii
Oleksii:
»
Skeuomorphism was widely popular in UI design for computers, mobile phones and then smart watches for a long time. Probably the most famous example is Apple UI under Steve Jobs leadership. The purpose of skeuomorphism was to make interfaces “affordable” and easily understandable by users who didn’t have prior experience of interacting with new devices.
Over time, it became clear that people got used to the virtual interfaces and there was no more need in direct representation of physical things on our screens, so the design became flat and minimalistic.
…But yesterday I saw a post on Threads by @yasirbugra [showing the idea of a virtual watch appearing on your wrist when you turned it as if to read the time].
It came to my mind that this is the real possibility of seeing skeuomorphism back, maybe even in a more functional form than we had on flat screens! If AR/XR devices like Quest, Vision Pro and their successors become popular among wide audience, it might be easier and more convenient for people to start using them with UI that closely resembles real objects, and not flat screens hanging around. This is not the case for any usage obviously, but I can imagine that “leaving” virtual “things” (like a notebook or a book) on a table or a bookshelf can be a nice step of making AR blend with our daily life seamlessly. And unlike for the case with 2D screens, we can now interact with virtual object in different ways by moving them around, rotating etc.
«
What’s wrong with this rental listing? The furniture is AI • Vice
Hayden Vernon:
»
Michael Anthony wasn’t sure straightaway, but it was the kid’s bunk bed that gave it away. The bed’s two overlapping ladders were a design decision that no human would make, but the sort of uncanny slip that has become the hallmark of AI generated images.
“I got an inkling as I went through the pictures, and suddenly it just didn’t seem right,” the 28-year-old software engineer tells VICE. Originally, he thought the weird pictures in the Rightmove listing were just a case of an over-eager estate agent on Photoshop. But after posting some of the images on the Spotted on Rightmove subreddit, people pointed out it was more than likely AI. “To be honest, I thought it was pretty impressive,” he says. “It’s come a long way in the past year really, hasn’t it?”
Michael had stumbled across a growing phenomenon: estate agents furnishing listing photos with AI. Agents are increasingly playing their own surreal version of The Sims, dressing up adverts with AI furniture to make depressing rental flats look like they’ve come out of the IKEA catalogue.
Virtual staging, as the process of adding fake furniture to property pics is known, has been around for a while, but it was previously done using standard image editing software. With the growth of generative AI tools, agents no longer have to rely on expensive graphics people (or their own crap Photoshop skills) to furnish property adverts.
«
Wonder what the expensive graphics people are going to do now. Estate agents (realtors, in the US) are always quick to embrace The New Thing: they have managed to find a real use for drones, taking aerial pictures of properties. Jumping onto the AI bandwagon is all in a day’s work.
unique link to this extract
Streaming services are spoiling the Super Bowl • The Verge
Jacob Kastrenakes:
»
Tens of millions of people gathered together to watch the Super Bowl last night — but they weren’t all watching it at exactly the same time. Cable streams of the game delivered footage that was around 50 seconds behind what was happening on the field, and the figure for online streaming services was even worse. Some viewers were watching a stream that was a minute-and-a-half behind the real world, leaving plenty of time for social media posts and push notifications to spoil what was about to happen in the game.
The figures, from streaming tech company Phenix, show that streaming still has a ways to go to catch up with other — and, often, older — methods of watching TV. Hulu, NFL Plus, and DirecTV Stream were on average more than a minute behind the action on the field. Fubo TV was the worst, with an average delay of almost 87 seconds and some users seeing a delay of up to two minutes.
There’s always going to be some delay between a real-time event and the footage appearing on a TV set across the country. But Phenix’s data shows that there’s very much room for improvement. Verizon FiOS had a delay of just 29 seconds. And the best performance of all came from the oldest method out there: over-the-air broadcasts. People getting the game from a broadcast signal experienced the briefest delay, seeing what was happening just 22 seconds after it occurred on the field.
These delays may sound brief, but they make a big difference in the way people experience live events today. It may take a minute for video to get from the field to your TV, but a social media post about a touchdown or interception can make it to your phone much quicker. The delay doesn’t matter if you’re just watching with friends, but nowadays, just about every viewer has access to faster sources of information in their pocket — the delay both puts them behind the online conversation and reveals what’s coming next.
«
This does raise the question of how the social media posts are faster than the video stream. Is the implication that people at the game are posting updates? Or just people on OTA broadcasts?
unique link to this extract
Falsehoods programmers believe about time zones • Zain Rizvi
Zain Rivzi:
»
My aunt has a problem. She loves joining Zoom meetings, but they’re all hosted in different time zones. It’s hard to remember if she should add 4 hours, subtract 3, or what. She’s not the most technical person, so google isn’t an option. She has to ask for help.
Every. Single. Time.
And, for the less technically minded, it’s also error-prone.
It got me thinking: What if event organizers could share a link that would do the work for you? If someone clicked on mytime.at/5pm/EST, they would see their local version of that time. It sounded simple enough.
I began coding. I knew trying to manage time is a fool’s errand, but that’s what datetime libraries are for. I would merely build an extra time zone conversion layer on top. Surely that couldn’t be complicated.
…Right?
I soon discovered just how wrong I was. One after another, I kept learning the falsehood of yet another “fact” that had seemed obviously true. Eventually my original vision became literally impossible to pull off without making serious compromises (more about that in a future blog post).
Hopefully this list will help you avoid the landmines I stepped on. All the falsehoods below are ones I’d considered true at some point in my adult life.
Most of them I believed just one month ago.
«
Why dynamic message signs confuse many drivers and increase the number of accidents • Frequent Business Traveler
Jonathan Spira:
»
The driver seeks out information and services which he thinks he needs; others, he tends to ignore. Adding extraneous and superfluous messages about sleighs and Yule can only serve to confuse the already overloaded driver.
Only when the severe changes or unusual conditions that require abrupt adjustments in maneuvering the vehicle appear does the driver require proper, timely and attention getting warnings.
A good example of such a situation occurred on December 28, 2023, when 15 states issued “heavy fog” warnings. Variable-message signs warned of “Heavy Fog” and told drivers to “Reduce Speed.” Another occurred on January 16, 2024, when 100 million people in the United States remained under windchill advisories and winter storm warnings, and signs were programmed to warn of “Poor Driving Conditions,” telling drivers to “Reduce Speed.”
While messages such as these demonstrate not only the proper and intended use of variable-message signage but also will cause minimal information overload, attempts to successfully display longer instructive or educational messages continue to elude the custodians of the signs.
Drivers in New York State saw a message that reads “Crash. No Injuries. Minor Damage.” We undertook an informal survey of 30 licensed drivers and asked them what the meaning of this particular sign was. Over 80% said that they understood it to mean that there was a minor crash ahead, albeit with no injuries and minor damage.” That, in fact, was my initial interpretation as well… until I kept seeing it on almost every sign.
My research, which entailed calling the state’s Department of Transportation, revealed that the sign was intended to inform drivers that, should they happen upon the scene of an accident, the new New York State “Move Over” law requires them to slow down and move left one full lane (the same goes for other stopped vehicles and roadway construction crews, incidentally).
Despite the fact that this sign is almost universally misunderstood, it continues appear throughout the New York metro area, leaving driver after driver with the impression that he was approaching the scene of an accident.
«
Later in the piece it mentions “humourous” signs, but the study outcomes on these are conflicting. What gets my goat is dynamic signs which show information that ceased to be correct hours ago, and which the controllers should know is wrong because they have traffic cameras.
unique link to this extract
His best friend was a warthog. One day, it decided to kill him • Texas Monthly
Peter Holley:
»
as he sat in a pool of his own blood on a beautiful October evening in 2022, Austin Riley couldn’t help but acknowledge the morbid absurdity of his current predicament. He’d spent decades conquering brain injuries only to be killed while doing mundane chores on his family’s 130-acre Hill Country ranch in Boerne. “After all I’d been through,” he said, “I just couldn’t believe that this was how it was going to end.”
As he slumped against a fence and his mangled body began to shut down, Austin’s mind went into overdrive. He thought about his girlfriend, Kennedy, whom he’d never get a chance to marry, and the children he’d never be able to raise. He thought about how much he loved his parents and how badly he wished he could thank them for the life they’d provided. He thought about the land before him, a valley accentuated by crimson and amber foliage that seemed to glitter in the evening light, and realized it had never seemed more beautiful than it did in that moment.
«
This is not a short read. But it is worthwhile. The description of what happened to him, and what he had to do in order to survive, is incredible. It also shows how you get a clarity of thinking when you’re about to die from calamitous injuries.
unique link to this extract
Prompt Engineering is a job of the past • We Are Developers
Adrien Book:
»
For a long time, my thinking on prompt engineering was as follows:
• Productized LLMs are a brand-new technology that people have not yet fully tamed
• Twitter reply guys have just been burnt by crypto and need a new grift
• They see AI emerge as the “next big thing”, but have no engineering or coding talent; they do, however, speak English (barely)
• They rebrand themselves as “prompt engineers” (sounds fancy! Looks good on LinkedIn!) and share obvious advice to “help” people get “the most out of ChatGPT”If it doesn’t sound like a real job that someone might be paid to do, trust your instinct. The above is correct for a majority of the “prompt engineers” you see on social media. However, having spoken to actual experts, my thinking has evolved. The truth is closer to:
• Productized LLMs are a brand-new technology that people have not yet fully tamed
• Sometimes LLMs behave in unexpected ways, and we need to understand why
• Because AI is a black box, we also need to understand what it can and cannot do so it can be better marketed
• Doing so involves more data analysis than spending 8 hours a day making wild guesses into a text box… but that’s also part of the jobThough this role is indeed important for today’s tech company, it was more so two years ago. Prompt engineering is doomed to disappear in the coming months.
…Less than a year ago, as MidJourney was taking off, there were talks of people making a living by selling elaborate prompt. How quickly this became laughable! It’s like calling yourself a typist in 2020. Technology has evolved in such a way that everyone can do it; it’s becoming less of a distinct profession and more a skill integrated into broader roles.
The irony is not lost on me that the first job created by AI might also be the first to vanish. Although the role will remain lucrative for a select few data scientists ($300k+)… the whole profession is an outlier. In fact, over the next few months, we will see “real” prompt engineers pivot towards becoming solutions engineer and work closely to clients.
«
Canada to ban the Flipper Zero to stop surge in car thefts • BleepingComputer
Sergiu Gatlan:
»
The Canadian government plans to ban the Flipper Zero and similar devices after tagging them as tools thieves can use to steal cars.
The Flipper Zero is a portable and programmable pen-testing tool that helps experiment with and debug various hardware and digital devices over multiple protocols, including RFID, radio, NFC, infrared, and Bluetooth.
Users have been demonstrating Flipper Zero’s features in videos shared online since its release, showcasing its capacity to conduct replay attacks to unlock cars, open garage doors, activate doorbells, and clone various digital keys.
“Criminals have been using sophisticated tools to steal cars. And Canadians are rightfully worried,” Canadian Industry Minister François-Philippe Champagne tweeted on Wednesday. “Today, I announced we are banning the importation, sale and use of consumer hacking devices, like flippers, used to commit these crimes.”
…According to the Canadian government, around 90,000 vehicles (or one car every six minutes) are reported stolen every year, with car theft resulting in $1bn in annual losses, including insurance costs for fixing and replacing stolen cars. The figures shared by the Canadian government when describing the car theft surge currently impacting Canada align with the most recent data shared by the Statistics Canada government agency, which shows an increasing number of car theft reports since 2021.
«
OK so it’s a scary terrible device, right? Not so fast:
»
Flipper Devices, the company behind the devices, says the gadget can’t be used to steal vehicles built within the last 24 years.
“Flipper Zero can’t be used to hijack any car, specifically the ones produced after the 1990s, since their security systems have rolling codes,” Flipper Devices COO Alex Kulagin told BleepingComputer.
«
Even in 2017, the average age of vehicles in Canada was 10 years. Stable/horse/door/bolted.
unique link to this extract
Could a rogue billionaire make a nuclear weapon? • WSJ
Sharon Weinberger:
»
The report [on this question, written by the Pentagon in 2013] read like background notes for an airport thriller: a Bond-villain-like corporation would set up shop as a legitimate business, managing a series of nominally independent subsidiaries responsible for different parts of weapons production in locations around the world. One company, for example, would be responsible for designing the centrifuges; another would produce the highly enriched uranium; a third would do the chemical processing. A company could even work directly with a rogue nuclear power. “Would our hypothesized enterprise ever go into partnership with North Korea? Or perhaps with Iran?” the report asks.
When I first read the study in 2018, it struck me as a fascinating premise, but implausible. Nuclear weapons are the domain of nations for a reason: they require huge facilities, big budgets and technical expertise you can’t exactly advertise for on LinkedIn.
But now, just five years later—and more than a decade after the study was completed—much has changed in the world. Private companies have long been involved in building weapons, including nuclear weapons, but the federal government has traditionally been the one funding and controlling the technologies. Now, even the Pentagon acknowledges that private capital is the dominant source of funding for key technologies.
…“What could you achieve with a billion? What could you achieve with $10 billion? And is that really beyond the reach of some individuals?” [report contributor Brian Jenkins] said. “Well, in the case of an Elon Musk, it means you can put rockets into space without NASA. And that’s the point of this.”
Miles Townes, who left graduate school to work on the study, told me he still thinks about it every day. He said the study was sparked by the way gas centrifuges had lowered the bar to entry when it came to nuclear weapons.
In the movies, spies turn over detailed designs for a nuclear weapon. But nuclear experts argue that what is truly secret is not necessarily how to design the weapon—much of that exists on the Internet or in the public domain—but the processes necessary to make the critical materials, like highly enriched uranium.
«
OK, perhaps let’s keep Musk just focussed on Twitter.
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
The Overspill: when there's more that I want to say 