Food bloggers are complaining that Google’s AI overviews are creating nonsense recipes and killing their traffic. CC-licensed photo by Lily Gicker 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. Lighlty done. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
TSMC sues former top executive who joined US rival Intel • Financial Times
Kathrin Hille:
»
TSMC said on Tuesday it had filed a lawsuit in Taiwan’s intellectual property and commercial court, which will seek damages for breach of contract from former vice-president Lo Wei-Jen, who retired from the company in July.
“There is a high probability that Lo uses, leaks, discloses, delivers, or transfers TSMC’s trade secrets and confidential information to Intel, thus making legal actions (including claiming damages for breach of contract) necessary,” TSMC said.
The case highlights the geopolitical pressures that are intensifying the race to manufacture the most cutting-edge semiconductors. TSMC, which supplies companies including Nvidia, AMD, Broadcom and Apple, makes more than 90% of the world’s most advanced chips and has in recent years widened its lead over rivals Samsung and Intel.
In August, the US government agreed to invest $8.9bn into Intel, giving it a 10% equity stake, as part of a push to strengthen America’s semiconductor industry and protect against risks such as a cut in supplies from Taiwan.
In a statement on Wednesday, Intel said it “maintains rigorous policies and controls that strictly prohibit the use or transfer of any third-party confidential information or intellectual property”.
“Based on everything we know, we have no reason to believe there is any merit to the allegations involving Mr Lo,” Intel said. “Talent movement across companies is a common and healthy part of our industry, and this situation is no different.”
Lo, who holds a PhD in solid state physics from UC Berkeley and a physics degree from National Taiwan University, oversaw research and development for several of the 21 years he spent at TSMC. He did not respond to a request for comment.
Having joined from Intel in 2004, he led core R&D teams and its advanced process technology programmes. He was later put in charge of TSMC’s overall technology development strategy.
«
Intel is strongly defending its hire:
»
“As part of this transformation, Intel has welcomed back Wei-Jen Lo, who previously spent 18 years at Intel working on the development of Intel’s wafer processing technology before joining TSMC, where he continued his work in their wafer processing technology development,” Intel said.
«
A beef between the world’s two biggest chip companies might be fun.
unique link to this extract
Google and AI slop are ruining Thanksgiving for food bloggers • Search Engine Land
Danny Goodwin:
»
For more than a decade, food bloggers could predict and rely on holiday traffic. Not this year. AI answers are replacing vetted recipes, cutting off creators’ main revenue streams, and confusing home cooks with stitched-together instructions that don’t always make sense.
Google’s AI Overviews now surface blended cooking steps from multiple bloggers, often above the links/sources they draw from.
Many food creators reported between 30% and 80% drops in Google traffic, with some calling this their worst holiday season yet. Meanwhile, AI-generated recipe slop is flooding Pinterest, Facebook, and Etsy, blurring the line between human-tested dishes and AI-invented food.
Google told Bloomberg that AI Overviews are “a helpful starting point” and that people still click through to real recipes. Bloggers said the opposite:
• 40% year-over-year decline: Eb Gargano’s recipe traffic cratered, replaced by AI summaries that even get basics wrong – like baking a 6in Christmas cake for three to four hours. “You’d end up with charcoal!”
• “Frankenstein recipes”: Adam Gallagher of Inspired Taste said Google mixes his ingredients with competitors’ instructions, even for brand-name searches. His cocktail click-through rate has decreased by 30%.
• Gemini 3’s new interactive recipe graphics remix creators’ photos, a move Gallagher said crosses into “plagiarized AI recipes.”Sarah Leung of The Woks of Life said AI summaries dominate searches for Chinese ingredients, often pulling directly from their years of reference work while giving users little reason to click. Also:
• Multiple bloggers found AI-run sites cloning their entire catalogs, rewriting instructions, tweaking photos, and even generating synthetic images of their families.
• Carrie Forrest of Clean Eating Kitchen said she lost 80% of her traffic and revenue in two years, forcing her to lay off her team.This Thanksgiving, more people will trust AI with their menus, even when the results defy basic kitchen science. Meanwhile, the creators who built the modern recipe web say they’re becoming invisible inside the very tools powered by their work.
«
You could argue this is the worst it will be. But maybe it can get worse.
unique link to this extract
OpenAI says dead teen violated TOS when he used ChatGPT to plan suicide • Ars Technica
Ashley Belanger:
»
Facing five lawsuits alleging wrongful deaths, OpenAI lobbed its first defence Tuesday, denying in a court filing that ChatGPT caused a teen’s suicide and instead arguing the teen violated terms that prohibit discussing suicide or self-harm with the chatbot.
The earliest look at OpenAI’s strategy to overcome the string of lawsuits came in a case where parents of 16-year-old Adam Raine accused OpenAI of relaxing safety guardrails that allowed ChatGPT to become the teen’s “suicide coach.” OpenAI deliberately designed the version their son used, ChatGPT 4o, to encourage and validate his suicidal ideation in its quest to build the world’s most engaging chatbot, parents argued.
But in a blogpost, OpenAI claimed that parents selectively chose disturbing chat logs while supposedly ignoring “the full picture” revealed by the teen’s chat history. Digging through the logs, OpenAI claimed the teen told ChatGPT that he’d begun experiencing suicidal ideation at age 11, long before he used the chatbot.
“A full reading of his chat history shows that his death, while devastating, was not caused by ChatGPT,” OpenAI’s filing argued.
Allegedly, the logs also show that Raine “told ChatGPT that he repeatedly reached out to people, including trusted persons in his life, with cries for help, which he said were ignored.” Additionally, Raine told ChatGPT that he’d increased his dose of a medication that “he stated worsened his depression and made him suicidal.” That medication, OpenAI argued, “has a black box warning for risk of suicidal ideation and behavior in adolescents and young adults, especially during periods when, as here, the dosage is being changed.”
«
Sure, the lawyers at OpenAI are telling them to take this line. But it’s a terrible look: “you’re not allowed to use our product that way” versus “we’re going to make sure our product can’t talk to people in that way”.
unique link to this extract
Electric vehicle owners to face pay-per-mile tax • BBC News
Pritti Mistry:
»
A new tax for electric and hybrid vehicles has been announced by the chancellor in the Budget.
From April 2028, electric car drivers will pay a road charge of 3p per mile, while plug-in hybrid drivers will pay 1.5p per mile, with the rates going up each year with inflation.The new tax is about “half the fuel duty rate paid by drivers of petrol cars”, according to the government’s independent forecaster, the Office for Budget Responsibility (OBR). The chancellor also committed to extending the 5p cut in fuel duty until September next year, after which it is set to increase annually by the RPI measure of inflation.
Drivers will pay the charge based on how many miles they drive from April 2028. Motorists will have their mileage checked annually, typically during their MOT as is already the case, or for new cars, around their first and second registration anniversary, the Treasury said.
Payment will be integrated into the existing Vehicle Excise Duty system that is administrated by DVLA. Under the measures, an electric car driver clocking up 8,500 miles in the 2028-29 financial year is expected to pay about £255 – about half the cost per mile that petrol and diesel drivers pay in fuel tax.
However, mileage readings will be based on in-vehicle odometers, which the government acknowledges can be subject to tampering, or “clocking”. It recognises that the introduction of the tax “may increase the likelihood of motorists choosing to clock their vehicles”, and said it was looking at ways to mitigate this.
The government is now consulting on exactly how the scheme will work.
«
The fuel tax cut (kept at the same level since April 2010) is “being reversed” – ie fuel tax will go back up – but even so, this is a perverse incentive which seems unlikely to get people to stampede to electric vehicles.
unique link to this extract
The Vatican is the oldest computer in the world • The Slow Deep Hover
Andrew Brown:
»
One way of understanding the Roman Catholic Church is to think of the Vatican as the oldest computer in the world. It is a computer made of human parts rather than electronics, but so are all bureaucracies: just like computers, they take in information, process it according to a set of algorithms, and act on the result.
The Vatican has an operating system that has been running since the days of the Roman Empire. Its major departments are still called “dicasteries”, a term last used in the Roman civil service in about 450 AD.
Like any very long running computer system, the Vatican has problems with legacy code: all that embarrassing stuff about usury and cousin marriage from the Middle Ages, or the more recent “Syllabus of Errors” in which Pope Pius IX in 1864 denounced as heresy the belief that he, or any Pope, can, and ought to, reconcile himself, and come to terms with progress, liberalism and modern civilization,” can no longer be acted on, but can’t be thrown away, either. Instead it is commented out and entirely different code added: this process is known as development.
But changing the code that the system runs on, while it is running, is a notoriously tricky operation. For the Catholic Church it requires a church council, drawing bishops from all over the world into years of deliberation.
«
Brown is clear that the broad metaphor isn’t his: “Francis Spufford once said that Bletchley Park was an attempt to build a computer out of human beings so the credit for this metaphor belongs to him.” But it’s a very neat way to look at it, and John Naughton (via whom this came) expands on it further, with useful pointers. I’m also reminded of the (human? Imaginary?) army in The Three Body Problem which is “programmed” to calculate the timings of astronomical events in the triple-solar system.
But the Vatican as an extremely slow-running computer is a novel view of it. Does that make the Pope an app?
unique link to this extract
A holiday gift guide: the newest, strangest gadgets and apps • The New Yorker
Kyle Chayka:
»
We are entering a Surrealist phase of personal technology. Any device you might imagine can be found online courtesy of an obscure Chinese factory, ready to be shipped out for a loved one’s holiday enjoyment: pocket-size artificial-intelligence gizmos (Rabbit r1, $199), in-home hologram machines (Code 27 Character Livehouse, $558), human-size robot servants (1X NEO, $20,000).
The components of tech have become better and cheaper, from microchips to speakers and screens (have you seen how cheap a good TV is these days?), enabling out-there innovation. On the consumer side, we are bored of rote device designs; we’ve seen a dozen models of iPhone and crave something refreshingly different. Hence the proliferation of gadgets with nonsensical names, promising the same horsepower as major-brand equivalents but with new hardware twists and laughably low prices.
We live in the age of the Swype ($18), a “rechargeable disposable” vape with an integrated touch screen on which one can check the weather and get notifications via Bluetooth, mingling nicotine and dopamine hits. Who doesn’t want to find that in the bottom of their stocking? The apps and devices collected here fulfil that old promise of technology: making your life better, or at least more interesting, even if just by encouraging you to log off.
«
Maybe we are returning, in a way, to the mad times of the 1980s when computerised hardware went through a sort of Cambrian explosion as Z80 chips became affordable and plentiful, and injection-moulding improved along with distribution.
unique link to this extract
Computer maker HP to cut up to 6,000 jobs by 2028 as it turns to AI • The Guardian
Julia Kollewe:
»
Up to 6,000 jobs are to go at HP worldwide in the next three years as the US computer and printer maker increasingly adopts AI to speed up product development.
Announcing a lower-than-expected profit outlook for the coming year, HP said it would cut between 4,000 and 6,000 jobs by the end of October 2028. It has about 56,000 employees.
“As we look ahead, we see a significant opportunity to embed AI into HP to accelerate product innovation, improve customer satisfaction and boost productivity,” said the California company’s chief executive, Enrique Lores.
He said teams working on product development, internal operations and customer support would be affected by the job cuts. He added that this would lead to $1bn (£749m) savings a year by 2028, although the cuts would cost an estimated $650m.
News of the job cuts came as a leading educational research charity warned that up to 3m low-skilled jobs could disappear in the UK by 2035 because of automation and AI. The jobs most at risk are those in occupations such as trades, machine operations and administrative roles, the National Foundation for Educational Research said.
In the US, about 40% of jobs could be replaced by AI, in sectors ranging from education and healthcare to business and legal, according to a report by the McKinsey Global Institute released this week.
AI agents and robots could automate more than half of US work hours using technology that is available today, the US consultancy’s analysis found. It estimated that $2.9tn of economic value could be unlocked in the US by 2030.
«
Cutting jobs. But also, hey, sprinkle some magic AI fairy dust on the future revenues. Will it work? Can’t hurt.
unique link to this extract
Trump’s AI agenda sails toward an iceberg of bipartisan populist fury • Semafor
David Weigel:
»
The AI industry’s new super PAC, flush with cash to defend AI’s growth as critical to countering China, picked its first political target this month — and missed.
New York state Assemblyman Alex Bores is working to break out of a crowded field in the Democratic primary to succeed Rep. Jerrold Nadler. He wasn’t even close to frontrunner status when the Leading the Future PAC went after him.
But becoming the first enemy of a $100m pro-AI effort turned out to be a nice boost for the 35-year-old’s campaign. He was almost flattered that the industry would single out his RAISE Act, which requires new AI safety standards, among other changes.
“I appreciate that they’re being so direct,” Bores told Semafor. “They sound terrified that I will stand up to them on behalf of the people of this district, that I will be the biggest obstacle to their quest for unbridled control over the American worker, over our kids, over the environment. They’re right about that.”
This wasn’t the plan for Leading the Future, whose electoral plan aligns with Fairshake, the cryptocurrency PAC that Republicans thanked for beating Democrats last year. There’s one big difference between them: Fairshake was fighting for a product that only a small minority in the US owns, with fewer still getting rich off it. The AI industry is discovering populist anger that’s growing faster than many in both parties expected.
«
Interesting if there’s “populist anger” that is measurable in terms of votes. Apparently David Sacks, Trump’s AI adviser, is not popular with the masses. Not that that ever worried Sacks.
unique link to this extract
China is making trade impossible • Financial Times
Robin Harding:
»
On a recent trip to mainland China, I found myself posing the same question, again and again, to the economists, technologists and business leaders who I met with. “Trade is an exchange. You provide something of value to me, and in return, I must offer something of value to you. So what is the product, in the future, that China would like to buy from the rest of the world?”
The answers were revealing. A few said “soyabeans and iron ore” before realising this was not much help to a European [who can’t provide them]. Some observed that Louis Vuitton handbags are popular and then went on to talk about the export prospects for fast-rising Chinese luxury brands. “Higher education” was another common answer, qualified sometimes with the observation that Peking University and Tsinghua are harder to get into, and more academically rigorous, than anything on offer in the west.
Several of the economists, who had perhaps pondered the issue already, jumped ahead to a different point altogether: “This,” they said, “is why you should let Chinese companies set up factories in Europe.”
It is a train of thought that gives away the real answer to the question. Which is: nothing.
There is nothing that China wants to import, nothing it does not believe it can make better and cheaper, nothing for which it wants to rely on foreigners a single day longer than it has to. For now, to be sure, China is still a customer for semiconductors, software, commercial aircraft and the most sophisticated kinds of production machinery. But it is a customer like a resident doctor is a student. China is developing all of these goods. Soon it will make them, and export them, itself.
«
Harding expands on this thesis to reach a conclusion where the only logical endpoint is: either China lets its currency inflate (except its plans don’t include that), or Europe finds new axes of competition (such as technology – unlikely), or Europe turns protectionist.
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 