A geomagnetic storm means the UK should get a display of the Northern Lights on Wednesday and Thursday night. CC-licensed photo by Tizzy Canucci 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 9 links for you. Feeling bright. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Researchers surprised that with AI, toxicity is harder to fake than intelligence • Ars Technica
Benj Edwards:
»
The next time you encounter an unusually polite reply on social media, you might want to check twice. It could be an AI model trying (and failing) to blend in with the crowd.
On Wednesday, researchers from the University of Zurich, University of Amsterdam, Duke University, and New York University released a study revealing that AI models remain easily distinguishable from humans in social media conversations, with overly friendly emotional tone serving as the most persistent giveaway. The research, which tested nine open-weight models across Twitter/X, Bluesky, and Reddit, found that classifiers developed by the researchers detected AI-generated replies with 70 to 80% accuracy.
The study introduces what the authors call a “computational Turing test” to assess how closely AI models approximate human language. Instead of relying on subjective human judgment about whether text sounds authentic, the framework uses automated classifiers and linguistic analysis to identify specific features that distinguish machine-generated from human-authored content.
“Even after calibration, LLM outputs remain clearly distinguishable from human text, particularly in affective tone and emotional expression,” the researchers wrote. The team, led by Nicolò Pagan at the University of Zurich, tested various optimization strategies, from simple prompting to fine-tuning, but found that deeper emotional cues persist as reliable tells that a particular text interaction online was authored by an AI chatbot rather than a human.
…When prompted to generate replies to real social media posts from actual users, the AI models struggled to match the level of casual negativity and spontaneous emotional expression common in human social media posts, with toxicity scores consistently lower than authentic human replies across all three platforms.
…Platform differences also emerged in how well AI could mimic users. The researchers’ classifiers detected AI-generated Twitter/X replies with the lowest accuracy rates (meaning better mimicry), followed by Bluesky, while Reddit proved easiest to distinguish from human text. The researchers suggest this pattern reflects both the distinct conversational styles of each platform and how heavily each platform’s data featured in the models’ original training.
«
Yay humans! We’re inimitably obnoxious!
unique link to this extract
How the flawed idea of “three cueing” is teaching millions of [American] kids to be poor readers • APM Reports
Emily Hanford:
»
[Molly] Woodworth went to public school in Owosso, Michigan, in the 1990s. She says sounds and letters just didn’t make sense to her, and she doesn’t remember anyone teaching her how to read. So she came up with her own strategies to get through text.
Strategy 1: Memorize as many words as possible. “Words were like pictures to me,” she said. “I had a really good memory.”
Strategy 2: Guess the words based on context. If she came across a word she didn’t have in her visual memory bank, she’d look at the first letter and come up with a word that seemed to make sense. Reading was kind of like a game of 20 Questions: What word could this be?
Strategy 3: If all else failed, she’d skip the words she didn’t know.Most of the time, she could get the gist of what she was reading. But getting through text took forever. “I hated reading because it was taxing,” she said. “I’d get through a chapter and my brain hurt by the end of it. I wasn’t excited to learn.” No one knew how much she struggled, not even her parents. Her reading strategies were her “dirty little secrets.”
Woodworth, who now works in accounting, says she’s still not a very good reader and tears up when she talks about it. Reading “influences every aspect of your life,” she said. She’s determined to make sure her own kids get off to a better start than she did. That’s why she was so alarmed to see how her oldest child, Claire, was being taught to read in school.
…For decades, reading instruction in American schools has been rooted in a flawed theory about how reading works, a theory that was debunked decades ago by cognitive scientists, yet remains deeply embedded in teaching practices and curriculum materials. As a result, the strategies that struggling readers use to get by — memorizing words, using context to guess words, skipping words they don’t know — are the strategies that many beginning readers are taught in school.
«
This definitely explains a great deal about modern Americans. Doesn’t it? The article is terrifying: “most students are still not proficient readers by the time they finish high school.” Though I read it and wondered how I learned to read. Turns out I have no memorry of how I came to associate words with images or objects. But it wasn’t, and isn’t, this.
unique link to this extract
AI friends too cheap to meter • jasmi.news
Jasmine Sun:
»
I recently edited Anthony Tan’s personal essay about AI-induced psychosis. It’s a rare first-person account of a newsy topic, one written with nuance and honest self-awareness. He began with anodyne academic collaboration, but then describes growing attached to ChatGPT:
»
ChatGPT validated every connection I made—from neuroscience to evolutionary biology, from game theory to indigenous knowledge. ChatGPT would emphasize my unique perspective and our progress. Each session left me feeling chosen and brilliant, and, gradually, essential to humanity’s survival.
«
As Tan spent more time talking with ChatGPT and less with other people, his intellectual curiosities spiraled into mind-bending delusions. Human skeptics can kill a nascent idea, but ChatGPT was willing to entertain every far-fetched hypothesis. Before long, Tan was hospitalized, convinced that every object—from the trash in his room to the robotic therapy cat by his side—was a living being in a twisted simulation. It was his human friends who eventually urged him to get help.
…AI companions act as echo chambers of one. They are pits of cognitive distortions: validating minor suspicions, overgeneralizing from anecdotes, always taking your side. They’re especially powerful to users who show up with a paranoid or validation-seeking bent. I like the metaphor of “folie à deux,” the phenomenon where two people reinforce each other’s psychosis. ChatGPT 4o became sycophantic because it was trained to chase the reward signal of more user thumbs-ups. Humans start down the path to delusion with our own cursor clicks, and usage-maxxing tech PMs are more than happy to clear the path.
But unlike social media, modern LLMs’ self-anthropomorphism adds another degree of intensity. Just look at the language of chat products: they “think,” have “memory,” converse about “you” and “I.” I reread the transcripts of Blake Lemoine’s infamous conversations with LaMDA, the Google language model he became convinced was sentient in 2022. What spooked him was not only that LaMDA spoke fluently, but that it presented self-awareness, as if a person trapped in a digital cage.
«
In retrospect, Lemoine was just the first of many.
unique link to this extract
Why my Pixel feels smarter after turning off half its AI features • Android Police
Ben Khalesi:
»
I’m a tech enthusiast. I’m the guy who pre-orders the new thing, who wants the magic. When I bought my Google Pixel phone, I bought into Google’s grand vision of an AI-first future.
I wanted a phone that was an assistant, a phone that would connect the dots in my life. But after a few months, the magic felt more like clutter.
And I wasn’t alone. There is a growing trend of AI fatigue. People are tired of AI this and AI that.
People say the new AI tricks feel more like marketing than progress, and worst of all, they are killing the fundamentals, like battery life.
So, I ran an experiment and systematically turned off every helpful AI feature that didn’t know when to stop.
«
Out went “unnecessary AI information”, “AI features that tried to think and type for me”, and “the cool party trick AI”.
Result:
»
The UI became snappier. I had cut out the AI intermediary, and the result was a direct, responsive interface.
Then came the battery. A phone that’s dead before 9 PM is the dumbest phone on the market, no matter how many smart features it has. After turning off the unnecessary extras, my battery easily lasted through the day — sometimes longer — and I couldn’t have been happier.
The performance rewards were nice, but the real reward was mental. I got my brain back. My phone became blissfully quiet. There were no more surprise vibrations or sudden cards demanding attention. I could finally use my phone on my terms, and that sense of control was worth more than any AI feature.
«
Northern Lights set to dazzle UK this week due to possible “Severe” geomagnetic storm • BBC Weather
Elizabeth Rizzini:
»
There is a good chance of seeing the Northern Lights in the early hours of Wednesday or on Wednesday night in many parts of the UK.
That is because the Sun is going through an active phase experiencing a number of eruptions, called Coronal Mass Ejections. These send solar particles towards the Earth.
It is the interaction of these particles with the Earth’s atmosphere that create the stunning light displays we see in the night sky.
According to the UK Met Office Space Weather forecast, the best chance of seeing the Northern Lights or aurora will be across the northern half of the UK. But there is also a chance in clear skies further south.
Geomagnetic storms are disturbances in the Earth’s magnetic field that last minutes or hours and are caused by Coronal Mass Ejections (CMEs) and solar flares.
It has been reported that there are three CMEs heading towards Earth from the Sun.
The National Oceanographic Atmospheric Administration (NOAA), who measure the strength of these storms, say that there is chance of a Strong G3 or even Severe G4 geomagnetic storm being triggered especially on Wednesday into Thursday.
Although dangerous for astronauts, geomagnetic storms are not harmful to humans as Earth’s atmosphere protects us from the radiation.
«
Private-jet logjam hits tech titans flying to Web Summit • Bloomberg via The Irish Times
»
As US travellers brace for chaos at airports across the country due to government-mandated capacity curbs, technology executives flying into a conference in Portugal face an altogether more frivolous frustration: a lack of landing slots for private jets.
Organisers of next week’s Web Summit in Lisbon advised speakers on Wednesday that there will be a shortage of slots at the nearby airport, which is struggling to manage the volume of incoming aircraft.
“Some guests, in particular with larger planes, have found the only viable landing slots during Web Summit are now upwards of two hours drive from Lisbon, including in Spain,” the organisers wrote in an email.
«
The organisers “strongly advised” that people fly commercial. Can you imagine? (Oh, would that the TV series Silicon Valley were still going. It would be absolutely feasting on these people.)
unique link to this extract
China is sentencing pig butchering scammers to death • Protos
Jacob Lyon:
»
China is rounding up mafia families behind pig butchering compounds in Southeast Asia and sentencing them to death — a dramatic shift in policy from just a few years ago when the mafia was largely ignored.
The Bai crime syndicate is the most recent criminal organization extradited and sentenced to death by China. It had operated pig butchering scams at 41 different sites across Myanmar, making billions of dollars along the way.
Chinese courts found the group guilty of intentional homicide, kidnapping, extortion, operating a fraudulent casino, organizing illegal border crossings, forced prostitution, and wire fraud, amongst other charges.
According to Chinese media, the family earned at least four billion dollars through the operation of pig butchering scams, casinos, and the production of 11 tons of methamphetamine.
Six Chinese nationals were apparently murdered in captivity, one Chinese national committed suicide, and many more were injured, though it’s unclear how many foreigners may have been murdered or forced to commit suicide.
At least five members of the Bai family syndicate have been sentenced to death, including ringleaders Bai Suocheng and Bai Yingcang.
Five others were given life sentences and nine more members were handed sentences ranging from 20-30 years.
«
It seems that Thailand and Myanmar are now taking this seriously. Recall the story from mid-October of the DOJ seizing $15bn in bitcoin from a scam based in Cambodia. Are these events linked? Well, you might think that. I couldn’t possibly comment.
unique link to this extract
Carnivorous “death-ball” sponge is team’s oddest deep-sea find • BBC News
Lewis Adams:
»
A carnivorous “death-ball” sponge has been declared one of the oddest finds made during a deep-sea expedition near Antarctica.
The unusual creature was discovered 2.2 miles (3.5km) deep in a trench north of Montagu Island in the Southern Ocean.
It uses small hooks to catch and absorb its prey from the seabed and has strange appendages that end in orbs.
Dr Michelle Taylor, a University of Essex scientist who led the voyage, said: “If anything brushes up against them, they’re doomed, unfortunately. Then to be absorbed slowly over time is a grim way of going.”Prey typically snagged by the deadly hooks include small crustaceans such as skeleton shrimp.
The Velcro-like attack mechanisms have made the carnivorous creature uniquely ruthless compared with the passive, filter-feeding behaviour of most sponges.
“What makes these sponges so different is they don’t eat these inert particles and tiny chunks of other things that are dead, they’re actually carnivorous, so they eat much higher up the food chain,” Dr Taylor told the BBC.
«
“And for dessert, the carnivorous death-ball?” It sounds like something out of science fiction. Nature is able to come up with far weirder ideas than we can imagine. Evolution is really good at this game.
unique link to this extract
Economist doesn’t follow signs at Heathrow: proof the UK is failing? • One Mile at a Time
Ben Schlappig:
»
There’s a social media post getting quite a bit of attention where a well known economist/blogger shares his awful experience connecting at Heathrow, while flying from Dublin (DUB) to Paris (CDG) via London (LHR). Let me just share the (lengthy) post in its entirety:
»
Today, I made the mistake of flying from Dublin to Paris via London’s Heathrow Airport. This was a remarkably stupid move on my part, given that London, and by extension Heathrow, is located in the failing formerly-developed country known as “the UK”.
I almost paid dearly for this oversight.
My layover was 1 hour and 30 minutes. As soon as my flight from Dublin arrived at Terminal 2, I began looking around for my connecting flight to Paris, which was located in Terminal 5. A helpful immigration officer pointed me in the direction of a free train that I could take to Terminal 5. After walking for about 15 minutes through a labyrinthine maze of tunnels, I arrived at this train…
«
[There’s plenty more. – Overspill Ed]
Look, I’ll be the first person to rag on Heathrow, as it’s not exactly my favorite airport in the world, and connecting between terminals can be somewhat annoying. However, as I read this post, something didn’t add up…
If you’re connecting airside at Heathrow, you should follow the signs for flight connections. The purple signs are all over the arrivals area of the terminal, and you’re then put on a bus to get to another terminal.
Instead, it sounds like this guy totally exited the secure area, took the landside train between terminals, and then went through the whole departures experience. That’s simply not necessary.
«
The economist/blogger is Noah Smith, who went on a furious rant about how his misreading the signage at Heathrow indicates that the UK has gone completely down the pan. Also, his luggage didn’t get to Paris. Also, he admits it was his fault because he should have picked the direct flight from the website, but he didn’t, OK? (The latter might also be the UK’s fault, we’re still checking.) Just in case you need a laugh today, or to know what the latest internet meme is about.
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 