So you’ve got a USB-C cable – but how do you find out what (if any) data speed it runs at? A new Mac app will tell you. CC-licensed photo by HS You 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. Charging ahead. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Uber’s autonomous vehicle strategy: slow their adoption • WIRED
Aarian Marshall:
»
A DECADE AGO, then-Uber CEO Travis Kalanick said he saw autonomous vehicles as an existential threat to the ride-hail company’s business model.
“What would happen if we weren’t a part of that future? If we weren’t part of the autonomy thing? Then the future passes us by,” Kalanick told Business Insider.
In the years since, Uber has settled on a strategy that, rather than see it build and operate its own self-driving cars, puts it on track to become the place where riders can get connected with any ride, driven by a human or robot. “We think there are going to be many AV players around the world, and we want to be the go-to commercial platform for all of them,” now-CEO Dara Khosrowshahi told investors in 2024. Since then, the company has signed agreements with more than 25 major robotaxi players, with driverless vehicles from Waymo, Nuro, Baidu, and Volkswagen’s MOIA either available or soon to be available on the Uber app in several global cities.
Now, according to documents viewed by WIRED and another obtained through a public records request, Uber’s lobbyists are pushing to build that strategy into law. The company’s representatives have pressed lawmakers to deploy autonomous vehicles on what it calls “hybrid networks,” where human drivers work alongside robots as the new tech grows.
In New Jersey, a lobbyist representing Uber took the strategy a step further, circulating legislative language that would, for a period of three years, require any platform offering driverless ride-hailing services to have human drivers serve 85% of its rides.
The language would likely prevent self-driving vehicle developers, including Waymo, Zoox, and Tesla, from operating their own ride-hail apps in the state—effectively forcing them onto another ride-hail app if they hope to enter the market and limiting competition for Uber, the country’s reigning ride-hail leader.
«
Very sneaky. A comparison would be insisting that any electric vehicle has to have a petrol engine, just in case, thus blocking EV entrants from entering the market. But Uber is very flexible: it’ll change its principles any time, as long as they suit it.
unique link to this extract
Meta pulls new AI image feature after days of backlash • BBC News
Kali Hays:
»
Meta has abruptly taken down a new feature that allowed people to use its artificial intelligence (AI) tool to make fake images from user content on Instagram.
The feature was part of a broad rollout of Muse Image, a new AI image generation tool Instagram’s parent company released last Tuesday.
It allowed users of the Meta AI chatbot to tag public-facing accounts on Instagram and quickly use content on those accounts to create AI-generated or altered content and images.
The feature quickly sparked blowback due to privacy concerns, leading Meta to admit it had “missed the mark” so it was “no longer available”.
Muse Image was the tech firm’s first foray into AI image generation but faced backlash as Instagram users were opted in by default. It meant that anyone with a public account could have their likeness used without their knowledge or permission.
Hollywood union Sag-Aftra described the U-turn as a “win”. It had previously urged its members and “all Instagram users” to take action to protect their likeness stating that there had been an “utter miscalculation of public sentiment regarding the obvious dangers and harms inherent in such use”.
The London-based human rights charity Privacy International had also criticised the feature, telling the BBC it was “the latest sign AI companies see people’s images and data as raw material to be exploited”.
«
Well that was predictably quick, wasn’t it. Though the usual playbook from Meta (and Google) is to quietly bring it back in almost exactly the same form, with some minor tweaks, but the same level of privacy, ie none.
unique link to this extract
Publishers are preparing to opt out of Google Search • Adweek
Mark Stenberg:
»
Beginning Sept. 15, all new websites signing up for Cloudflare, as well as all the customers on its free tier, will have the default settings in their bot management protocol set to block “multi-purpose crawlers” on any webpage that has ads. This means that any crawler that scrapes for both search indexing and AI training will be turned away at the door, unless the site owner decides otherwise.
“We’ve been clear about what we want,” said Cloudflare chief strategy officer Stephanie Cohen. “We want a technical solution that allows you to be discoverable without having to give your content away for free.”
While a handful of crawlers fit this description—Apple and Bing, among others—the primary, unnamed target of this action is Google, which infamously uses one crawler to both index sites and train its AI models.
In doing so, Google forces publishers to make an impossible choice: They either allow both functions, enabling Google to scrape their content to train the AI products that are regurgitating their data without compensation; or they shut off both functions and disappear from Google Search, presumably losing their largest source of traffic in the process.
To be fair, Google recently introduced an option called Google Extended, which nominally allows publishers to opt out of AI training without disappearing from Search. But publishers are wary that the program will penalize their search visibility, according to executives at two media companies.
…Historically, abandoning Google Search would have been commercial suicide, according to SEO consultant Lily Ray. It is simply too valuable of a source of audience discovery and traffic.
“It’s a really hard tradeoff. Some publishers have already blocked OpenAI until it strikes partnerships, but with Google it’s hard,” Ray said. “Google is a different conversation because it has so many more users than other AI firms.”
But the gradual erosion of search traffic in recent years has, paradoxically, given publishers more agency to consider walking away from the platform.
USA Today Inc., which encompasses not just USA Today but a nationwide network of news sites, is weighing its options on the matter, according to CEO Mike Reed.
The company, like many others, has responded to declines in search traffic by bolstering audience from other sources, like newsletters, social media, and events. Its traffic has remained relatively stable in recent years, hitting its goal of 1 billion pageviews every month for the last three years, according to Reed.
«
Where to find the colors your screen can’t show you • Ryan Moulton
Ryan Moulton takes you through how the eye works, how computer screens work, and why what the eye sees goes beyond what the screen can show:
»
When you look at a plant under normal light, its leaves are almost always within the sRGB triangle. Plants are green, but they aren’t that green. Their leaves absorb a lot of blue and red light, but not so much that it pushes us to the edge of the colorspace. The magic happens in a deciduous forest, when the light isn’t just reflected, it is transmitted. The transmittance curves of foliage are much more selective than their reflectance curves, so the color you see passing through a leaf is much more saturated than the color that bounces off of it. You’ve probably noticed this in person. A leaf lit by sunlight looks from the top to be relatively ordinary, but from underneath, it glows.
A single pass through a leaf knocks out all of the blues, and half of the reds, but the light then continues on, passing through other leaves, and bouncing off other leaves. These effects stack exponentially. The more times the light interacts with a leaf, the more it is purified to its spectral peak, generally around 550 nm. The colors you’ll see will be all the greens and yellows contained in the lobe traced out by the paths of repeated reflections and repeated transmissions. A green leaf lit by light that passes through another leaf one time is already outside of the gamut, greener than green.
When you’re standing in a maple forest at noon in the middle of summer, the intensity of the green is indescribable. Being in a fully lit and fully leafed deciduous forest is like being underwater if the water were green, which brings us to our next subject, water.
Water aggressively absorbs reds, slowly absorbs greens, and barely absorbs blues at all. This pattern pushes nearly any spectrum with blue and green in it out of the sRGB gamut almost immediately. When you look at sand in the shallow water near the coast, it traces a curve through colorspace as the depth of the water changes. The light from the sun is filtered once as it passes through the water on the way down, bounces off the sand, and filtered again as it comes back up to your eye. White or yellow sand will first shift to unrepresentable cyans, then to unrepresentable blues, and then finally converges close to the sRGB blue primary again once the water is very deep and dark.
«
Intel to invest €5bn in Irish plant as AI chip demand surges • Financial Times
Barbara Moens, Tim Bradshaw and Jude Webber:
»
Intel will invest €5bn to expand semiconductor manufacturing in Ireland as the US chipmaker seeks to grow production of processors used in AI systems, in a boost for Europe’s ambitions to build a stronger chip industry.
The company said on Monday the investment would upgrade its Leixlip facility west of Dublin, expanding current production and advancing research and development.
The announcement comes as demand for AI data-centre chips continues to surge, prompting semiconductor companies to increase manufacturing capacity. The EU is pursuing its Chips Act to boost manufacturing capacity in Europe and reduce its reliance on foreign suppliers of high-performance chips.
Intel’s executive vice-president Naga Chandrasekaran said the investment ensured “that Ireland remains at the forefront of the world’s most advanced manufacturing ecosystems, while strengthening the region’s role in the global technology landscape”.
«
According to Intel’s page about Leixlip, the company has so far invested over €30bn in the facility, so this is quite a substantial addition; €17bn of that was put in from early 2019.
From teetering on the edge to dropping billions in foreign countries: Intel really has been saved by the AI bubble.
unique link to this extract
PC market runs low on RAM to grow: shipments decline 4.9% as memory crunch bites • IDC
»
Worldwide PC shipments fell 4.9% year over year in Q2 2026 to 68.2 million units, according to IDC — the first decline after nine consecutive quarters of growth. A persistent memory chip shortage drove the reversal, pushing vendors to pull inventory forward as far as possible.
Beyond the memory supply crunch, other components such as storage, along with geopolitical issues, have continued to weigh on the market.
“The real story here is the disconnect between units and dollars: shipments are falling, but revenue is climbing because vendors are pushing through price increases faster than demand is dropping,” said Jitesh Ubrani, research director for consumer devices at IDC. “Given worsening macro conditions and a memory shortage that isn’t expected to ease until early 2028, we don’t expect another round of inventory pull-forward, which points to a sharp slowdown in growth rates in the second half of 2026. Vendors are bracing for further price hikes into 2027, and channels are already flagging concern about elevated inventory at these higher price points.”
«
Top five: Lenovo, HP, Dell, Apple, ASUS. Amazingly, Apple actually increased sales (from 6.1m to 6.7m, nearly 10%) because of the low-cost Macbook Neo. Assuming of course that all the IDC figures are reliable, but we’ve had to go with that for a while.
As ever one tends to suspect that the low end market is going to struggle during this crisis, and lose share to the big ones. That is a cycle which keeps repeating, and somehow they keep hanging on, even while their share gets eroded slowly year by year.
unique link to this extract
This free Mac app reveals the truth about your mystery USB-C cables • The Verge
Sean Hollister:
»
Nearly three years ago, I showed you an awesome $8 cable tester that quickly tells you if your USB-C cable is likely fast, slow, powerful, or weak. Sadly, that gadget got discontinued, and I’ve never found anything as intuitive or inexpensive since. But if you’ve got a Mac with Apple Silicon chips, you can simply download an even more impressive tester for free.
It’s called WhatCable, and it works by reading the data your Mac already collects about attached USB devices, data that Apple doesn’t normally pass along to you. Just click a little widget that lives in the menu bar atop your Mac, and you can see every USB-C cable and device attached to your computer.
Here’s how creator Darryl Morley explained it to me:
»
Every Apple Silicon Mac has a port controller chip that handles USB Power Delivery negotiation. When you plug in a cable with an e-marker, the port controller sends a “Discover Identity” message to the chip in the cable and gets back a structured message: vendor ID, speed rating, current rating, voltage limits, whether it’s active or passive, and so on.
macOS writes that response into the IOKit registry. WhatCable reads it using Apple’s public APIs. No root access, no private entitlements. The data isn’t hidden, Apple’s firmware does the negotiation and publishes the result. It’s just not surfaced anywhere in standard macOS tooling. WhatCable reads what’s already there.
The e-marker is one source. WhatCable also reads from the Mac’s own hardware, the actual negotiated connection speed, Thunderbolt link speed, and live voltage and current at each port. The connected device tells us what it is, who made it, and what it supports. Put all three together, cable, device, and Mac, and WhatCable can tell you not just what everything claims to support, but what’s actually happening on the connection right now, and which part is the bottleneck if something isn’t performing as expected.
«
«
Immediate download! Though USB data cables are quite hard to come by. Power cables, no problem. The difficulty is identifying the ones that will actually carry data.
unique link to this extract
The most famous AI writing tic is also the most mysterious • The Atlantic
Will Oremus:
»
If Julius Caesar had debuted this year, William Shakespeare might have been accused of writing it with AI. A certain suspicious rhetorical device appears again and again in the play. It’s in Act I, Scene ii: “The fault, dear Brutus, is not in our stars, but in ourselves.” In Act III, Scene ii: “Not that I loved Caesar less, but that I loved Rome more.” And later in that same scene: “I come to bury Caesar, not to praise him.”
These famous lines include what has become perhaps the best-known tic of AI writing—a sentence that tells you what the subject isn’t as well as what it is: It’s not X; it’s Y. Once you start noticing the construction, you see it all over the place. In one version, the Y is additive: It focuses, intensifies, or expands on the X. An annual review by Citizens Financial Group reported that growth in its private-banking division was “not just a win for the private bank—it’s a win for the entire enterprise.” In another variant, the Y supplants the X as the preferred descriptor. “The target was never a man. The target was the truth,” Michael Flynn, a former Donald Trump adviser, wrote in a March X post.
…Although chatbots have advanced dramatically in their research and reasoning capacities, they are still fundamentally text-prediction machines. They generate answers one “token”—or chunk of text—at a time, based on what has come before. Each successive word choice factors in both the statistical likelihood of that word coming next in a sequence, based on patterns in the original training data, and the likelihood that it will lead to a highly rated response overall. In other words, the models are always seeking a balance between the clever word choice and the obvious one.
When a chatbot uses negative parallelism, according to this theory, it’s essentially hedging between the two. Once it has started a sentence whose function is to characterize something, the path of least resistance is to say first what the thing isn’t (X), and only then what the thing is (Y). Put another way: For a sentence that begins with “This is,” following it with “not just” is both more likely and safer than the many options for how to directly characterize its subject. And after “This is not just,” the rest of the sentence gets easier too. The next word can be X—the boring, obvious descriptor that gets negated—which in turn sets up the final choice of Y, the somewhat punchier descriptor.
Even if researchers could figure out exactly why chatbots embrace negative parallelism, there’s another factor that could make it very hard to fix: “When something gets into these models, it’s very hard to pull it out,” Masrour, the Pangram engineer, said.
«
It’s not good – it’s annoying.
unique link to this extract
Older adults in China know AI is slop. They just like it • Rest of World
Viola Zhou:
»
Young people can’t seem to stop watching AI slop videos of cats talking and fruits cheating on each other. Older people are enjoying a different kind of AI-generated content, which provides them with much-needed comfort and companionship.
Take Uncle Chang, a 67-year-old family friend who recently visited New York from Taiwan. As we chatted, the retired businessman showed me some YouTube videos that made him cry. In one, an AI-generated young blonde woman named Rose Bennett performs “Whiskey Was Louder Than Me,” a song about growing up with an alcoholic father after her mother passed away. In another, Rose sings “Brother Became My Father” together with her brother (also AI). Their AI father and the AI audience were in tears.
The videos reminded Chang of his own childhood. His mother, too, had left him, after suffering from his father’s beating. He was eventually raised by his older sisters. “To me, it was ‘sisters became my mother,’” Chang said. “These songs tell such touching stories.”
Chang’s experience reminded me of the “AI family” videos that are getting popular on Chinese social media. On TikTok-like platforms Douyin and Kuaishou, AI-generated chubby babies or handsome adult sons send daily blessings, tell viewers how much they miss them, and bring along virtual roses. Some AI influencers even take on the role of the elderly audience’s virtual lovers.
…researchers found that the AI characters offer what real-world family members often do not. They express love directly (rare in Chinese families), show a higher level of filial piety, and talk about health and historical topics relevant to older people. One viewer told researchers that AI content had touched on experiences from her youth that her actual children, who hadn’t lived through the same history, couldn’t relate to. The AI videos also use soundtracks of folk music familiar to the older generation.
«
This feels like a worrying portent for the future.
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 