Thefts of copper have risen dramatically in the US as prices rocket due to data centre demand. CC-licensed photo by Trafigura Images 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 9 links for you. Unbending. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Here comes Apple’s new Siri again • The Verge
Allison Johnson:
»
if such a thing as a race to an AI assistant exists, Apple is losing badly. Gemini is already doing things like ordering Ubers and DoorDashing teriyaki. It can look at your calendar and figure out when you should leave for the airport. Gemini won the race, fair and square.
But there’s also a growing distrust of AI, particularly from young people, and the better Gemini gets, the creepier it is. It has to be if it’s going to deliver on the promise of a truly helpful assistant. But wanting your AI assistant to anticipate your next move and actually watching it happen? Those are very different things. I willingly gave Gemini permission to access my Google Photos and Gmail, but it always makes my skin crawl hearing Gemini say my son’s name out loud. I test out a lot of this stuff as it becomes available — hazard of the job — but the public reaction when these kinds of features start trickling down to the mainstream will be very telling.
New New Siri will be built on top of Gemini in some fashion. Apple is no doubt paying handsomely for the privilege, but there’s a potential upside to being one step removed in this way. You know what company doesn’t have its name attached to a big, unpopular data center project? Apple. Google isn’t winning friends and influencing people by rushing to start massive construction projects in backyards across the country. Apple gets to keep its hands clean, even if its payments to Google are presumably being funneled toward the great data center buildout.
Then there’s the Copilot of it all; the AI-buttons-everywhere factor. Siri’s attempts to summarize messages are amusing and often annoying, but at least Siri isn’t all up in every one of my work documents begging to summarize it for me. On the other hand, you can’t open a Google app without coming face-to-face with a Gemini sparkle these days, and it risks getting real old, real fast.
Don’t get me wrong; I think Apple would love to put Siri to work writing my emails, perfecting my photos into “memories,” and talking me through the next steps to rehabilitate the dying plants in my yard. It’s just that Siri can’t really do any of that yet. When we meet this new Gemini-enhanced Siri, it’ll be telling to see where and how aggressively it surfaces.
«
Siri, to remind you, was launched in September 2011; and that was from a company Apple bought. So the idea has been around for about two decades. But internal resistance in Apple to LLMs meant they weren’t seen as important until ChatGPT reset everything. Mark Gurman says there was a big reset meeting in early 2025 (MacObserver writeup).
And there might be new hardware: M5 Macbook Pros, and – restrain your enthusiasm – updated HomePod minis and Apple TVs. Yes, I know!
unique link to this extract
Brazilian airline Azul plans further frequency cuts as fuel shock bites • Reuters
Gabriel Araujo:
»
Brazilian airline Azul is stepping up capacity cuts amid higher jet fuel prices linked to the Iran war, and the carrier will continue to trim flying to protect cash in an uncertain environment, CEO John Rodgerson said.
Rodgerson told Reuters the industry’s largest companies were reducing capacity to better align with demand at higher cost levels, and Azul would follow suit, going beyond earlier cuts as the conflict drags on.
“When we made our initial cuts, we thought the war would be over by now,” he said in an interview on Friday, in the build-up to a meeting of global airline chiefs in Rio de Janeiro. “But it’s continuing, so we’re going to continue to opportunistically cut some frequencies, make sure that we’re only flying things that make sense.”
Most of Azul’s reductions in the second quarter were on international routes, with further adjustments focused on domestic frequencies rather than pulling entire cities, Rodgerson said. “Do you fly to Curitiba six times a day? Maybe with these fuel prices, it should be four.”
The airline was prioritizing its main hubs in Campinas, Belo Horizonte and Recife, he added. “We’re yet to pull cities, but that’s always on the table. But you first start with utilization and cutting frequencies. You don’t want to be utilizing an aircraft 13, 14 hours a day when fuel prices double.”
«
Smaller carriers and/or those in Asia might be among the first to feel this, but while the war (now 100 days old, allegedly “won” for about 94 of them) continues this is just going to spread.
unique link to this extract
Reliability of LLMs as medical assistants for the general public: a randomized preregistered study • Nature Medicine
Andrew Bean, Adam Mahdi et al (mainly at the Oxford Internet Institute):
»
We tested whether LLMs can assist members of the public in identifying underlying conditions and choosing a course of action (disposition) in ten medical scenarios in a controlled study with 1,298 participants. Participants were randomly assigned to receive assistance from an LLM (GPT-4o, Llama 3, Command R+) or a source of their choice (control).
Tested alone, LLMs complete the scenarios accurately, correctly identifying conditions in 94.9% of cases and disposition in 56.3% on average. However, participants using the same LLMs identified relevant conditions in fewer than 34.5% of cases and disposition in fewer than 44.2%, both no better than the control group. We identify user interactions as a challenge to the deployment of LLMs for medical advice.
Standard benchmarks for medical knowledge and simulated patient interactions do not predict the failures we find with human participants. Moving forward, we recommend systematic human user testing to evaluate interactive capabilities before public deployments in healthcare.
«
In other words: chatbots are good at diagnosis, but unskilled humans are bad at understanding what it’s telling them. (Thanks Joe S for the link.)
unique link to this extract
Amid a flood of AI advances, astrophysicists are questioning the soul of their field • Science
Joshua Sokol:
»
One afternoon in April, Cecilia Garraffo settled down at the head of a conference room table in Cambridge, Massachusetts, and gazed out at what might be the last astrophysicists of their kind.
The walls of this room had, in the past, reverberated with the din of thousands of other groups of scientists. Now, as streaks of sunlight poured in, the discussions turned to nonhuman collaborators. One by one, the gathered researchers discussed how they planned to apply machine learning to problems in astronomy. Observing an interstellar comet. Discerning wispy filaments of galaxies at the universe’s largest scales. Developing a new “tokenizer” that can translate astrophysical images into a form more readable by artificial intelligence (AI). “Sometimes models will be overconfident,” Garraffo warned a junior team member.
Afterward, as everyone filed out, black hole researcher Daniel Palumbo made a brief announcement. Representatives from AI chipmaker NVIDIA were on campus in search of scientists who wanted to solve problems using their hardware. To anyone who might need extra processing power, “today’s the day,” he said.
…Garraffo’s colleague Alyssa Goodman showed me a data-fitting problem. She wanted to understand how the spiral arms of a distant galaxy were moving. But isolating just that motion from other patterns imparted into her data by the spin and the geometry of that distant galaxy had thwarted her group for years. She asked ChatGPT, which resolved the problem in a few minutes. Now, her research group was planning to write several papers on the resulting data set, “the single best map of spiral arm kinematics ever—like, by a factor of 100.”
Conversations comparing these tools with human researchers, once an underground whisper, have grown into a deep rumble in astrophysics departments around the world. Many are turning over aspects of their research practice—searching the literature, developing code, writing proposals to use telescopes, doing first-pass “reads” of their peers’ submitted proposals, and actually solving problems—to agentic AI systems such as Anthropic’s Claude or OpenAI’s Codex.
…Already, by making it faster and easier to produce professional-seeming papers, AIs threaten both to overwhelm journals and peer reviewers and to take opportunities away from junior scientists. But far upstream of that, many scientists interviewed by Science sense a phase change underway. Many fear that if unleashed in all parts of the scientific process, AI tools could lead to nothing less than the death of astrophysics as a human endeavor. “A lot of people think that it’s too late to intervene—we’re done,” says David Hogg, a computational astrophysicist at New York University (NYU).
«
Singularities inside black holes are truly unavoidable • Big Think
Ethan Siegel:
»
inside of a black hole, the math of general relativity is very clear: all of that matter and energy that goes into forming it, no matter how it’s initially configured, is going to wind up collapsed down to either a single, zero-dimensional point (if there’s no net angular momentum) or stretched out into an infinitely thin one-dimensional ring (if there is “spin,” or angular momentum, present). Comedian Steven Wright even jokingly said, “Black holes are where God divided by zero,” and in some sense, that appears to be true: our calculations from all of our theories of physics, including general relativity and quantum field theory, break down under the conditions of a singularity. If physics doesn’t even make sense at that scale, how can we trust it to be correct about a singularity’s very existence?
While many hope that quantum gravity will save us from the inevitability of a singularity, many don’t think that even that is possible, for very good reasons. Here’s why a singularity at the center of every black hole ought to be completely unavoidable, no matter what sort of quantum theory of gravity happens to exist.
In principle, as Einstein first realized, if all you have is some configuration of matter that starts off distributed over some volume (with no rotation, no initial motions, and with space itself not expanding or contracting), the outcome is always the same: gravitational attraction will bring all of that matter together until it collapses down to a single point. Around that point, dependent on how much mass/energy there is all together, there will form a region of space known as an event horizon: a volume from within which the escape velocity, or the speed you’d need to travel to escape from this object’s gravitational pull, would be greater than the speed of light.
«
Well, that wraps it up. It’s an interesting article, though. (What the hell is actually happening in the singularity remains.. unknown.)
unique link to this extract
Comparisons: as Predictable as… the Sunrise • The Pudding
The Pudding:
»
An analysis of 200,000 similes from popular fiction.
«
That’s what it is! And there’s a lot to delve into. I hit the second most common simile in the first challenge. There’s plenty more where that came from, of course. (Via Helen Lewis’s Bluestocking.)
unique link to this extract
BBC director-general Matt Brittin: ‘It’s worth fighting for’ • Financial Times
Daniel Thomas lunches with the new director-general of the BBC, Matt Brittin, formerly CEO of Google UK:
»
The BBC has already struck a deal to post hundreds of hours of content on YouTube. Some executives worry that these sorts of deals just give away content to the benefit of the third-party platform.
“Ideally, you wouldn’t be starting from where we are, but look at the data. Young people are on YouTube five times as much as they’re consuming television, so we’ve got to be there. We need to have our content out in front of young audiences, whether it’s on TikTok or whatever,” he says. “We need to be closer, simpler and faster.”
It is hard to play in the centre . . . trying to be a media organisation that represents everyone and is trying to balance things
He is keen to improve the BBC’s technology, however, saying that past investment had been focused on content but “less focused on the product” such as the iPlayer service. “It hasn’t been able to accelerate as fast as it needs to. Could we leapfrog by partnering with others? We should be open to all these questions, because you’ve got to ultimately deliver value for the licence fee payer.”
Brittin was watching Mackenzie Crook’s comedy Small Prophets — “really good” — but was shocked to find there was no recommendation to another of his shows, The Detectorists. “That actually would have set me up for the bits of viewing I do with my wife on the sofa when we’re relaxing,” he says.
He would also like to see if the BBC could take a bigger role in areas such as education and skills — recalling how Google’s DeepMind team grew up in an era of the BBC Micro, a computer system developed in the early 1980s as part of a project that introduced a generation to coding — as well as in the development of UK-based large language models that could have “culturally sensitive and appropriate content”.
He has not thought through the details but says: “What if you took those ideals from the pioneers of the BBC about public service broadcasting and reinterpreted them today, then one key thing might well be around a British version of the AI revolution.”
«
The topic of licence fee renewal comes up: he says he can see the argument for a levy on streaming services, which might be a pointer towards how the future negotiations will go.
European publishers seek £552m+ from Google claiming ad market abuse • Press Gazette
Charlotte Tobitt:
»
More than 20 European news publishers are taking legal action against Google seeking damages of £550m for adtech monopoly abuses.
The case comes off the back of the European Commission handing Google a fine of €2.95bn (£2.55bn) last year for abusing its dominant position in online advertising technology.
The European Commission said that any people or company affected by anti-competitive behaviour outlined by this case could seek damages, which would be considered separately to the fine imposed on Google.
The publishers involved in the case argue they should collectively be awarded damages of more than €640m (£552m) due to the impact Google’s actions had on them. They believe they would have earned significantly higher advertising revenues and paid lower fees for adtech services if not for the fact Google had created a less competitive market.
Publishers are taking part from the Czech Republic, Estonia, France, Hungary, Finland, the Netherlands, Poland and Sweden. The case is being funded by Prague-based litigation funder LitFin, which will cover the costs even if it fails. The publishers involved have agreed to share part of any awarded damages with it if they win.
LitFin chief operating officer Matej Pardo said: “Google’s abuse of its position across the ad tech stack has been found unlawful at the highest levels – now it’s time for the publishers who bore the cost of that conduct to be made whole.
“By bringing a grouped claim, we can utilise efficiencies of scale to make this kind of action available to smaller players across Europe, who might otherwise not be in a position to bring a claim against such a deep-pocketed adversary as Google.”
…Other cases have previously been started against Google. In 2024 a coalition of 32 European media groups including Axel Springer and Schibsted brought a claim for €2.3bn (£2bn) alleging they suffered losses due to Google’s digital advertising practices.
Earlier this year five US publishers – Penske, The Atlantic, McClatchy, Conde Nast and Vox Media – sued Google alleging “deceptive and manipulative” adtech practices.
«
Google denies the claims. There are a lot of these lawsuits going on all over the world. (I’m a partner in one in the UK.)
unique link to this extract
Thieves are targeting copper wires. AT&T is fighting back • NPR
John Ruwitch:
»
Last year, AT&T recorded more than 10,400 incidents of copper wire theft nationwide — about 200 a week. Some 4,300 of those were in California.
It’s an outsized problem for the company, since only about 3% of AT&T’s customers are still connected by copper wire networks. That includes households without access to cell service or fiber optic connections, some in rural areas, as well as some businesses that still rely on old wires for internet or fax connections.
Copper cables, pioneered over 180 years ago by Samuel Morse, often hang next to their modern counterparts, fiber optic lines. When thieves cut cables for copper, they often slice fiber cables, too, because they look similar. A snipped fiber cable is what tripped this alarm.
Susan Santana, president of AT&T West, is on the ride-along. She says homes, hospitals, airports, schools and more can lose connections when cables are cut. The problem is “not an easy one to solve, by any means,” she says, but AT&T is trying. “We have locked down manhole lids with extra bolts. We’ve put sensors across our lines. In some instances we’ve had to hire private security guards,” Santana says.
AT&T has also offered a $20,000 bounty for information leading to the arrest and conviction of people involved in copper cable thefts.
The problem is not limited to telephone wires. Thieves have been cutting electric cables, too. The California city of San Jose has an online tracker for streetlights that have gone dark after being hit. EV chargers are also targets. Rick Wilmer, CEO of ChargePoint, the largest charging network in the United States, says it’s a problem they deal with every day. He says he got so frustrated that he started prototyping solutions on his own.
“I was so motivated I literally was going down to Home Depot and buying all kinds of different wire and Kevlar and stuff, and wrapping cables and taping it down and trying to cut it with my own pruning shears to see if it was, you know, making it any more difficult,” he says. He handed the project off to company engineers, who developed charger cables that are impregnated with cut-resistant material. The idea, he says, is that a thief might be able to hack off one of those wires, but their shears will be damaged in the process. They won’t be able to hit multiple chargers in one go.
«
Guess why prices of copper have rocketed? Rearrange these words for a clue: centres AI data.
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 