What if an MRI could be shrunk from a car-sized thing to a hand-sized one? One startup is trying. CC-licensed photo by Navy Medicine 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 10 links for you. Not that small. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
Alibaba’s Qwen AI model challenges US dominance despite chip restrictions • Rest Of World
Sam Eifling:
»
So far, the AI boom has been dominated by U.S. companies like OpenAI, Google, and Meta. In recent months, though, a new name has been popping up on benchmarking lists: Alibabai’s Qwen. Over the past few months, variants of Qwen have been topping the leaderboards of sites that measure an AI model’s performance.
“Qwen 72B is the king, and Chinese models are dominating,” Hugging Face CEO Clem Delangue wrote in June, after a Qwen-based model first rose to the top of his company’s Open LLM leaderboard.
It’s a surprising turnaround for the Chinese AI industry, which many thought was doomed by semiconductor restrictions and limitations on computing power. Qwen’s success is showing that China can compete with the world’s best AI models — raising serious questions about how long U.S. companies will continue to dominate the field. And by focusing on capabilities like language support, Qwen is breaking new ground on what an AI model can do — and who it can be built for.
Those capabilities have come as a surprise to many developers, even those working on Qwen itself. AI developer David Ng used Qwen to build the model that topped the Open LLM leaderboard. He’s built models using Meta and Google’s technology also but says Alibaba’s gave him the best results. “For some reason, it works best on the Chinese models,” he told Rest of World. “I don’t know why.”
In the short term, much of Qwen’s success comes from its unique position in the Chinese market. At launch, Alibaba claimed some 90,000 clients were using some models from Alibaba’s Tongyi Qianwen LLM series. (The name “Qwen” comes from a shortening of the term, which translates roughly to “all-encompassing knowledge.”) Most of the clients are Chinese companies that would be reluctant to form direct partnerships with U.S. companies like OpenAI or Anthropic.
«
World’s biggest banks pledge support for nuclear power • FT
Lee Harris and Malcolm Moore:
»
Fourteen of the world’s biggest banks and financial institutions are pledging to increase their support for nuclear energy, a move that governments and the industry hope will unlock finance for a new wave of nuclear power plants.
At an event on Monday in New York with White House climate policy adviser John Podesta, institutions including Bank of America, Barclays, BNP Paribas, Citi, Morgan Stanley and Goldman Sachs will say they support a goal first set out at the COP28 climate negotiations last year to triple the world’s nuclear energy capacity by 2050.
They will not spell out exactly what they would do, but nuclear experts said the public show of support was a long-awaited recognition that the sector had a critical role to play in the transition to low-carbon energy.
The difficulty and high cost of financing nuclear projects has been an obstacle to new plants and contributed to a significant slowdown in western countries since a wave of reactors was built in the 1970s and 1980s.
“This event is going to be a game-changer,” said George Borovas, head of the nuclear practice at law firm Hunton Andrews Kurth and a board member of the World Nuclear Association. Until now, he said, banks had found it politically difficult to support new nuclear projects, which often required sign-off from the chief executive’s office.
“Banks at their senior management level would just say, we don’t understand anything about nuclear. We just know it’s very difficult, very controversial.”
«
Promising; next you’d want the money to be withdrawn from funding fossil fuel projects.
unique link to this extract
Data center emissions probably 662% higher than big tech claims. Can it keep up the ruse? • The Guardian
Isabel O’Brien:
»
Big tech has made some big claims about greenhouse gas emissions in recent years. But as the rise of artificial intelligence creates ever bigger energy demands, it’s getting hard for the industry to hide the true costs of the data centers powering the tech revolution.
According to a Guardian analysis, from 2020 to 2022 the real emissions from the “in-house” or company-owned data centers of Google, Microsoft, Meta and Apple are probably about 662% – or 7.62 times – higher than officially reported.
Amazon is the largest emitter of the big five tech companies by a mile – the emissions of the second-largest emitter, Apple, were less than half of Amazon’s in 2022. However, Amazon has been kept out of the calculation above because its differing business model makes it difficult to isolate data center-specific emissions figures for the company.
As energy demands for these data centers grow, many are worried that carbon emissions will, too. The International Energy Agency stated that data centers already accounted for 1% to 1.5% of global electricity consumption in 2022 – and that was before the AI boom began with ChatGPT’s launch at the end of that year.
AI is far more energy-intensive on data centers than typical cloud-based applications. According to Goldman Sachs, a ChatGPT query needs nearly 10 times as much electricity to process as a Google search, and data center power demand will grow 160% by 2030. Goldman competitor Morgan Stanley’s research has made similar findings, projecting data center emissions globally to accumulate to 2.5bn metric tons of CO2 equivalent by 2030.
«
Hence, unsurprisingly, you might want to get a nuclear power station to be your provider – because, as Ben Thompson also points out, nuclear stations run 24/7.
unique link to this extract
Jony Ive confirms he’s working on a new device with OpenAI • The Verge
Alex Cranz:
»
Jony Ive has confirmed that he’s working with OpenAI CEO Sam Altman on an AI hardware project. The confirmation came today as part of a profile of Ive in The New York Times, nearly a year after the possibility of a collaboration between Altman and the longtime Apple designer was first reported on.
There aren’t a lot of details on the project. Ive reportedly met Altman through Brian Chesky, the CEO of Airbnb, and the venture is being funded by Ive and the Emerson Collective, Laurene Powell Jobs’ company. The Times reports it could raise $1 billion in funding by the end of the year but makes no mention of Masayoshi Son, the SoftBank CEO rumored last year to have invested $1 billion in the project.
The project only has 10 employees currently, but they include Tang Tan and Evans Hankey, two key people who worked with Ive on the iPhone. LoveFrom, Ive’s company, is leading the device’s design, according to the report. The team is reportedly now working out of a 32,000-square-foot office building in San Francisco, part of a $90 million strip of real estate that Ive has bought up on a single city block.
As for the device itself? The Times says that Ive and Altman discussed “how generative AI made it possible to create a new computing device because the technology could do more for users than traditional software” due to its ability to handle complicated requests. Last year, it was rumored to be inspired by touchscreen technology and the original iPhone.
But it sounds like few specifics are nailed down.
«
Guess: it will be a round flat oblong (perhaps with one side raised) with at most one button and one light. Perhaps a hidden speaker.
unique link to this extract
‘It’s not a solution for teen girls like me’: Instagram’s new under-18 rules met with skepticism • The Guardian
Alaina Demopoulos:
»
Meta, which owns Instagram, rolled out changes that give parents the ability to set daily time limits on the app and block teens from using Instagram at night. Parents can also see the accounts their children message, along with the content categories they view. Teen accounts are now private by default, and Meta said “sensitive content” – which could range from violence to influencers hawking plastic surgery – will be “limited”.
Teens with Instagram accounts will notice these rules go into effect within 60 days. If a child under the age of 16 wants to nix or alter these settings, they need parental permission; 16- and 17-year-olds can change the features without an adult. (One very easy loophole for teens: lying about their age. Meta also said it is working on improved age verification measures to prevent teens from circumventing age restrictions.)
“I feel these changes are very positive in a lot of ways, especially because they’re restricting sensitive content, but I don’t think it’s a solution,” Morton said. “Especially for teen girls, if you ask them what the main problem with Instagram is, they would say body image stuff.”
The issue of teen safety has dogged Meta since its start as Facebook, and these new rules come amid revived backlash from parents and watchdog groups. Instagram has come under fire for not protecting children from child predators and feeding them self-harm content. While testifying at a Senate hearing on online child safety in January, Meta CEO Mark Zuckerberg apologized to parents in the audience holding signs with pictures of children lost to suicide or exploited on the app.
And according to a 2021 Wall Street Journal investigation, researchers at Instagram have been studying how the app harms young users, especially young girls, for years.
«
Electronic warfare spooks airlines, pilots and air-safety officials • WSJ
Andrew Tangel and Drew FitzGerald; graphics by Adrienne Tong and Carl Churchill:
»
American Airlines Capt. Dan Carey knew his cockpit equipment was lying to him when an alert began blaring “pull up!” as his Boeing 777 passed over Pakistan in March—at an altitude of 32,000 feet, far above any terrain.
The warning stemmed from a kind of electronic warfare that hundreds of civilian pilots encounter each day: GPS spoofing. The alert turned out to be false but illustrated how fake signals that militaries use to ward off drones and missiles are also permeating growing numbers of commercial aircraft, including U.S. airlines’ international flights.
“It was concerning, but it wasn’t startling, because we were at cruise altitude,” Carey said. Had an engine failure or other in-flight emergency struck at the same time, though, the situation “could be extremely dangerous.”
Pilots, aviation-industry officials and regulators said spoofed Global Positioning System signals are spreading beyond active conflict zones near Ukraine and the Middle East, confusing cockpit navigation and safety systems and taxing pilots’ attention in commercial jets carrying passengers and cargo.
The attacks started affecting a large number of commercial flights about a year ago, pilots and aviation experts said. The number of flights affected daily has surged from a few dozen in February to more than 1,100 in August, according to analyses from SkAI Data Services and the Zurich University of Applied Sciences.
Modern airliners’ heavy reliance on GPS means that fake data can cascade through cockpit systems, creating glitches that last for a few minutes or an entire flight. Pilots have reported clocks resetting to earlier times, false warnings and misdirected flight paths, according to anonymized reports shared with government and industry groups.
«
(Thanks Karsten for the link.)
unique link to this extract
When you call a restaurant, you might be chatting with an AI host • Wired via Ars Technica
Flora Tsapovsky:
»
A pleasant female voice greets me over the phone. “Hi, I’m an assistant named Jasmine for Bodega,” the voice says. “How can I help?”
“Do you have patio seating,” I ask. Jasmine sounds a little sad as she tells me that unfortunately, the San Francisco–based Vietnamese restaurant doesn’t have outdoor seating. But her sadness isn’t the result of her having a bad day. Rather, her tone is a feature, a setting.
Jasmine is a member of a new, growing clan: the AI voice restaurant host. If you recently called up a restaurant in New York City, Miami, Atlanta, or San Francisco, chances are you have spoken to one of Jasmine’s polite, calculated competitors.
In the sea of AI voice assistants, hospitality phone agents haven’t been getting as much attention as consumer-based generative AI tools like Gemini Live and ChatGPT-4o. And yet, the niche is heating up, with multiple emerging startups vying for restaurant accounts across the US.
Last May, voice-ordering AI garnered much attention at the National Restaurant Association’s annual food show. Bodega, the high-end Vietnamese restaurant I called, used Maitre-D AI, which launched primarily in the Bay Area in 2024. Newo, another new startup, is currently rolling its software out at numerous Silicon Valley restaurants. One-year-old RestoHost is now answering calls at 150 restaurants in the Atlanta metro area, and Slang, a voice AI company that started focusing on restaurants exclusively during the COVID-19 pandemic and announced a $20m funding round in 2023, is gaining ground in the New York and Las Vegas markets.
«
And there you were thinking that it would be your Google phone that would be ringing up and making appointments. Well the joke’s on you, human.
unique link to this extract
Telegram changes policy, says it will provide user data to authorities • 404 Media
Joseph Cox:
»
Telegram updated its privacy policy on Monday to say that the company will provide user data, such as IP addresses and phone numbers, to law enforcement agencies in response to a valid legal order.
The news is a significant shift in Telegram’s policies on providing data to law enforcement. It comes after French authorities arrested Telegram’s CEO Pavel Durov in August, in part due to Telegram’s refusal to hand over data in response to lawful orders.
“If Telegram receives a valid order from the relevant judicial authorities that confirms you’re a suspect in a case involving criminal activities that violate the Telegram Terms of Service, we will perform a legal analysis of the request and may disclose your IP address and phone number to the relevant authorities,” the privacy policy read on Monday.
A day earlier, the policy only specifically mentioned terror cases. “If Telegram receives a court order that confirms you’re a terror suspect, we may disclose your IP address and phone number to the relevant authorities. So far, this has never happened,” an archived version of the policy reads.
«
Telegram (or more specifically Pavel Durov) deciding that keeping its bosses out of jail is more important than its users staying out of jail. Which makes sense, really.
unique link to this extract
Synex founder, once detained at the border with an 80-pound magnet, is building portable MRIs to test glucose • TechCrunch
Margaux MacColl:
»
Back in 2019, Synex Medical founder Ben Nashman spent the night detained by US customs. Nashman tried to explain he was simply transporting materials from Buffalo to Toronto for his homemade MRI. Customs, however, took issue with the label on the package: “nuclear magnetic resonance.”
Nashman spent hours in a bright waiting room before he finally convinced them that he was really just a run-of-the-mill 18-year-old scientist with an obsession with MRI technology. They let him take his roughly 80-pound magnet, and he zoomed back to Toronto. “I got back at like 3 or 4 am and got a few hours of sleep before classes,” he said.
Nashman, now 24, might have landed himself on a list of suspicious individuals, but he insists it was worth it: that one very long night was part of his years-long journey to build a portable MRI capable of testing glucose and other important molecules without the need to extract blood. Today, the company is one step closer to that goal, announcing a $21.8m Series A fundraise, with investors like Accomplice, Radical Ventures, Fundomo and Khosla Ventures. It brings the company’s total haul up to over $36m, with includes seed funding from Sam Altman.
Right now, Synex’s prototype is the size of a toaster, although Nashman hopes to one day have it fit in your palm. It works by first using MRI to create a 3D image of the finger to find the best spot to test. It then uses something called magnetic resonance spectroscopy to send radio pulses that “excite the different molecules,” Nashman said. The machine then takes the signals from all the molecules and filters for a specific one. Synex will start with glucose testing, but will eventually track things like amino acids, lactate and ketones.
«
Very ambitious project, but if it can be made to work, then it’s got a huge potential market for diabetics.
unique link to this extract
Sonos workers shed light on why the app update went so horribly • Ars Technica
Scharon Harding:
»
Employees Bloomberg spoke with claimed that internal frustrations were high close to the app’s scheduled launch date because the app wasn’t ready. Three former workers who recently worked there pointed to “yelling” and “screaming” in meetings as employees tried to warn higher-ups. Employees claimed that Sonos’ desire to get new customers and please investors was becoming more important than ensuring that old hardware would work properly with the new app. Indeed, this is exactly what happened when the app released. [CEO Patrick] Spence admitted in August that many customers, “especially those with some of our older products in their systems, are having an experience that is worse than” before. Examples included “existing speakers missing from their Sonos systems,” “errors while setting up new products,” and latency, per Spence.
“They thought they were making such a big, bold decision. It was the wrong decision,” someone Bloomberg described as a “former senior employee” said. Additionally, two people told Bloomberg that one former employee who worked on the app said during a meeting that they were worried about losing their job if they kept questioning the app’s release.
Per Bloomberg, Spence asked Sonos lead counsel Eddie Lazarus to investigate what led to the app debacle. Speaking with about 24 “key employees,” Bloomberg reported, Lazarus reported his findings to Sonos at the end of July.
Sonos’ head counsel said that Sonos created a list of “essential” bugs to fix before releasing the update but admitted: “Our list of essential bugs, obviously, was not comprehensive enough.”
Lazarus disagreed with the idea of a “breakdown in culture,” as Bloomberg put it, driving the bad app’s release. He noted that the app’s release was pushed from early 2024 until May in response to employee concerns.
In hindsight, it seems obvious that Sonos should have waited until the app was on par with the experience long-time users have enjoyed for years before rolling it out. It’s possible that it would have delayed the release of the Ace headphones, which were supposed to help Sonos revive its revenue after a post-pandemic decline. But now Sonos is delaying two product launches because of its hastiness.
«
The technical debt had built up literally for decades. But Sonos got the timing exactly wrong on when and how to pay it. Still a business school example of how not to do it.
| • 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 