Airlines in the US are becoming a lot more restrictive about what people can claim is “carry-on baggage”. CC-licensed photo by Bradley Gordon 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. They fit perfectly. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
Airlines are coming for your carry-on bags • WSJ
Dawn Gilbertson:
»
Fanny packs. Cross-body bags. Shopping bags. Pillows and blankets. The Southwest Airlines gate agent rattled off so many items that counted toward the two carry-on bag limit on my flight to Baltimore, I thought it might be a playful jab at Spirit and Frontier and their rigid carry-on policing to collect more fees.
But this was no joke. Southwest quietly began cracking down on carry-on bags on Feb. 22, ahead of the spring and summer travel rush, advising gate agents of the changes in a memo. This crackdown isn’t about bag size. It is about how many bags you have.
Southwest isn’t alone in putting passengers’ personal items in its crosshairs as a way to save precious bin space and speed up boarding. Delta and United agents have also recently asked me to stuff my small Lululemon bag in my backpack. One American Airlines frequent flier told me he watched gate agents in Sacramento, Calif., and Dallas list a litany of items that count as a personal item on weekend flights to Nashville, Tenn., last month.
Carting all your stuff to the gate can save you time and often saves money, especially with some airlines’ new, higher checked-baggage fees. Delta joined the club on Tuesday, announcing prices of $35 for your first bag and $45 for your second.
But testing airlines’ carry-on limits is now more likely to backfire, and lose you precious time as airlines make you consolidate items or check a bag at the gate. Few things sum up the industry’s carry-on challenges like Southwest’s latest move. The nation’s largest domestic carrier by passengers should have the fewest issues given its generous two-free-checked-bag policy. (Unlike checked bags, the government doesn’t track carry-on bag volume and airlines don’t disclose it.)
Southwest declined to discuss its carry-on changes beyond a statement saying the change “provides for a consistent customer experience and helps to align with other airlines’ policies.” A memo to employees about the changes singles out cross-body purses of any size and pillows and blankets, but employees are free to ad lib, spokesman Chris Perry says. Representatives for Delta, United and American pointed to their carry-on policies when asked for comment.
Tymali Gore, a traveling hospice nurse, couldn’t believe it when she heard a gate agent announce new rules about pillows, blankets and a host of other items counting as a personal item late last month. “It was the first time I’d ever heard anything like that,” she says.
«
Then again, some people pretty much bring a steamer trunk and try to wedge it into the overhead lockers, then give up and vainly attempt to stuff it under the seat in front. The mind boggles.
unique link to this extract
Inside the world of TikTok spammers and the AI tools that enable them • 404 Media
Jason Koebler:
»
We have recently been getting bombarded with Instagram Reels of influencers explaining how they make five figures a month by using AI to create tons of viral TikTok pages using stolen celebrity clips juxtaposed next to Minecraft gameplay footage. This strategy, the influencers say, allows them to passively make $10,000 a month by flooding social media platforms with stolen and low-effort clips while working from private helicopters, the beach, the ski slope, a park, etc.
What I found was a complex ecosystem of content parasitism, with thousands of people using a variety of AI tools to make low-quality spammy videos that recycle Reddit AMAs, weird “Would You Rather” games, AI-narrated “scary ocean” clips, ChatGPT-generated fun facts, slideshows of tweets, clips lifted from celebrities, YouTubers, and podcasts.
To help these people fill the internet with nonsense, there is an entire industry of creators, influencers, hustlers, and software developers selling them templates, stock clips, TikTok account creation services, cash out services, low-wage video editors in the developing world, AI voiceover and editing tools, and different “strategies” or “metas” to go viral enough to earn money from YouTube’s AdSense or from TikTok’s Creativity Program Beta, a monetization program that pays for “high-quality, longer TikTok videos” but which AI content influencers say can be easily gamed with low-effort content.
One of the kings of this world is Musa Mustafa, who got his start editing clips for the streamer Sneako but now seemingly makes most of his money from a Discord channel called “Media Metas,” which has 80,000 members and has a locked, premium section that costs $40 per month and is full of strategies and software people can supposedly use to go viral and make thousands of dollars a month. Whop, the platform he uses to sell access to the Discord, claims he is now making more than a million dollars a year through their platform.
«
Probably knock a couple of zeroes off that, but it’s slightly depressing that it’s an option at all.
unique link to this extract
European Electricity Review 2024 • Ember
Sara Brown and Dave Jones:
»
The EU accelerated its shift away from fossil fuels in 2023, with record falls in coal, gas and emissions. Fossil fuels dropped by a record 19% to their lowest ever level at less than one third of the EU’s electricity generation. Renewables rose to a record 44% share, surpassing 40% for the first time. Wind and solar continued to be the drivers of this renewables growth, producing a record 27% of EU electricity in 2023 and achieving their largest ever annual capacity additions. Furthermore, wind generation reached a major milestone, surpassing gas for the first time.
Clean generation reached more than two-thirds of EU electricity, double fossil’s share, as hydro rebounded and nuclear partially recovered from last year’s lows alongside the increase in wind and solar.
Coal was already in long-term decline, and that trend resumed in 2023. The temporary slowdown in coal plant closures during the energy crisis did not prevent a huge fall in coal generation this year, with a wave of plant closures imminent in 2024. Gas generation fell for the fourth consecutive year, and as coal nears phase-out in many countries, gas will be next to enter terminal decline.
In addition to clean growth, falling electricity demand also contributed to the drop in fossil fuel generation. Demand fell by 3.4% (-94 TWh) in 2023 compared to 2022, and was 6.4% (-186 TWh) lower than 2021 levels when the energy crisis began. This trajectory is unlikely to continue. With increased electrification, this rate of demand fall is not expected to be repeated in the coming years. To reduce fossil fuels at the speed required to hit EU climate goals, renewables will need to keep pace as demand increases.
«
That fall in electricity demand is peculiar: Ember puts it down to “a drop in industrial electricity consumption, mild weather and energy savings and efficiency” – principally in the energy-intensive industries of chemicals/petrochemicals, iron/steel, and paper/pulp, where manufacturing may have been reined in as gas prices soared.
unique link to this extract
Vendor offering citations for purchase is latest bad actor in scholarly publishing • Science
Katie Langin:
»
In 2023, a new Google Scholar profile appeared online featuring a researcher no one had ever heard of. Within a few months, the scientist, an expert in fake news, was listed by the scholarly database as their field’s 36th most cited researcher. They had an h-index of 19—meaning they’d published 19 academic articles that had been cited at least 19 times each. It was an impressive burst onto the academic publishing scene.
But none of it was legitimate. The researcher and their institution were fictional, created by researchers at New York University (NYU) Abu Dhabi who were probing shady publishing practices. The publications were written by ChatGPT. And the citation numbers were bogus: some came from the author excessively citing their own “work,” while 50 others had been purchased for $300 from a vendor offering a “citations booster service.”
“The capacity to purchase citations in bulk is a new and worrying development,” says Jennifer Byrne, a cancer researcher at the University of Sydney who has studied problematic publications in the biomedical literature. In academia, a researcher’s h-index and the number of citations they’ve garnered are often used for hiring and promotion decisions. And the fabricated profile, which was part of a study posted as a preprint on arXiv, shows “extreme” tactics that can be employed to manipulate them, adds Byrne, who was not involved in the work. (The researchers declined to name the vendor to avoid giving them more business.)
The study got started when Yasir Zaki, a computer scientist at NYU Abu Dhabi, and his colleagues noticed troubling patterns among real researchers. After combing through the Google Scholar profiles of more than 1.6 million scientists and looking at authors with at least 10 publications and 200 citations, the team identified 1016 scientists who had experienced a 10-fold increase in citations over a single year. “You know something is off when a scientist experiences a sudden and massive spike in their citations,” Zaki says.
«
Arguably this sort of thing would have been harder to spot in the days before Google Scholar – though there maybe wouldn’t have been the same incentive to do it.
unique link to this extract
Apple car’s crash: design details, Tim Cook’s indecision, failed Tesla deal • Bloomberg (archived)
Mark Gurman and Drake Bennett:
»
According to a longtime Apple executive who worked on the car, it was widely seen within the company as an ill-conceived product that needed to be put out of its misery. “The big arc was poor leadership that let the program linger, while everyone else in Apple was cringing,” they say. Asked what went wrong with the effort, a senior manager involved in the vehicle’s interior design replied: “What went right?”
…It was Steve Jobs who first floated the idea of a car at Apple. In the late 2000s, in a typically grand pronouncement, the company’s co-founder and CEO declared internally that Apple should have dominant technologies in all of the spaces in which people spent time: at home, at work and on the go. For many Americans, being in transit means being on the road, sometimes for hours a day. “We talked about what would be this generation’s new Volkswagen Beetle,” recalls Tony Fadell, who led mobile device engineering under Jobs. In the wake of the 2008 financial crisis, with American car companies on the brink of failure, the Apple chief executive even floated the idea of acquiring General Motors Co. for pennies on the dollar.
That scheme was quickly abandoned, in part because Apple decided it would be a bad look and in part because of the need to focus on the iPhone. But in 2014, seeking a new multi-hundred-billion-dollar revenue stream, Cook began to focus again on cars. Apple executives weighing whether to enter the market joked with one another that they’d rather take on Detroit than a fellow tech giant: “Would you rather compete against Samsung or General Motors?” The profit margins in cars were far lower than in consumer electronics, but Apple was coming off a stretch during which it had reshaped not only the music industry but the mobile phone market.
To its supporters, the idea of getting into automobiles had the potential to be, as one Apple executive puts it, “one more example of Apple entering a market very late and vanquishing it.” While the initial prototypes operated like traditional cars, these supporters eventually pursued more radical redesigns, invoking a transportation technology experience they said would “give people time back.” The ultimate plan was a living room on wheels where people who no longer needed to drive their cars could work or entertain themselves with Apple screens and services instead.
«
Absorbing read; more is going to come out about this. Lots of wrong choices and indecision.
unique link to this extract
Decoder guest host Hank Green makes Nilay Patel explain why websites have a future • The Verge
Hank Green, not of The Verge, interviewed Nilay Patel, editor-in-chief of The Verge, about how The Verge is still here:
»
HG: let’s start with you being the person who runs the last website on earth. Because you say things all the time and then you don’t explain them, which I love, but now I’ve got you. And so you have to explain to me why The Verge is “the last website on earth.”
NP: That’s a little bit of a joke. It’s 50% a joke. I’m aware that there are other websites. What I specifically mean is we were founded in a boom time of websites. We were founded in 2011. We started talking about the site in 2010. We remain part of a venture-backed digital media startup. There were a lot of those back then. We had a lot of competition in 2011, meaningful — like we were scared of them — competition.
ReadWriteWeb existed, and we tried to beat them every day. TechCrunch was a very different kind of publication back then. We tried to beat them all the time, and I really respect the people I competed against. I came up at Engadget competing ferociously against the people at Gizmodo, and we became first rivals and then really good friends out of that competition. Some of those sites still exist. Some of them are still doing great work. Some of them still have great people. But that moment when there was a ferocious rush of energy and money and attention into websites has obviously faded.
We’re not making those the same way we used to anymore, and I look at my peer group and so many of them are gone. To me, it’s that. It’s all the things: the people and the properties that I used to wake up in fear of, many of them are radically different than they used to be. And we’re still here. And that feels strange to me.
HG: It feels strange. You won, and it’s like, “Oh, I don’t actually…” It turns out that when you’re put into the arena and you’re the last man standing, there’s just a lot of carnage around, which isn’t that much of a triumph. It feels like it hurts a little bit. It’s weird to be us, our age, and hear that the word website feels almost anachronistic. It feels of another era.
The way I think about it is that I don’t have anyone else’s algorithm to think about, and that is really important to me. But then I look at all of the most important creators and the most influential members of the new media, and what they are is so successful that they have transcended algorithms on other people’s platforms.
«
The whole podcast is available to listen to (for free); there’s also the transcript. The Verge has indeed managed something remarkable in surviving and succeeding in its current form for so long.
unique link to this extract
Google hit with €2.1bn lawsuit from more than 30 European media companies • POLITICO
Pieter Haeck:
»
A group of 32 European media organizations have filed a lawsuit against Google, seeking damages of about €2.1bn.
The lawsuit touches on the US tech giant’s digital advertising practices, with the media groups claiming that they “incurred losses due to a less competitive market,” according to a statement shared by law firms Geradin Partners and Stek, which represent the organizations.
“Without Google’s abuse of its dominant position, the media companies would have received significantly higher revenues from advertising and paid lower fees for ad tech services,” the statement added.
Among the media groups are some of Europe’s leading news companies, including Axel Springer (owner of POLITICO), Norway-based Schibsted, and Benelux groups such as DPG Media and Mediahuis. The coalition claims to cover 17 European countries.
The lawsuit was filed in a Dutch court.
In June last year, the European Commission sent antitrust charges to Google over its advertising business.
“Our preliminary concern is that Google may have used its market position to favour its own intermediation services,” Executive Vice President Margrethe Vestager said at the time.
“Not only did this possibly harm Google’s competitors but also publishers’ interests, while also increasing advertisers’ costs.”The European Union’s competition watchdog has been probing Google’s online display advertising business since 2021. It’s previously probed the company’s shopping search service, its mobile phone software and advertising contracts, levying more than €8bn in fines.
«
(I’m involved with a similar lawsuit against Google in the UK; the process is ongoing.)
unique link to this extract
Amazon just bought a 100% nuclear-powered data center • Electrek
Michelle Lewis:
»
One of the US’s largest nuclear power plants will directly power cloud service provider Amazon Web Services’ new data centre.
Power provider Talen Energy sold its data center campus, Cumulus Data Assets, to Amazon Web Services for $650m. Amazon will develop an up to 960-megawatt (MW) data center at the Salem Township site in Luzerne County, Pennsylvania.
The 1,200-acre campus is directly powered by an adjacent 2.5 gigawatt (GW) nuclear power station also owned by Talen Energy.
The 1,075-acre Susquehanna Steam Electric Station is the sixth-largest nuclear power plant in the US. It’s been online since 1983 and produces 63m kWh per day. The plant has two General Electric boiling water reactors within a Mark II containment building that are licensed through 2042 and 2044.
According to Talen Energy’s investor presentation, it will supply fixed-price nuclear power to Amazon’s new data center as it’s built. Amazon has minimum contractual power commitments that ramp up in 120 MW increments over several years. The cloud service giant has a one-time option to cap commitments at 480 MW and two 10-year extension options tied to nuclear license renewals.
«
Not sure how I feel about this: OK, so the centre is going to be nuclear-powered: hooray. But isn’t that energy that could be used to power homes or other businesses? The tradeoff implied here is tricky.
unique link to this extract
Meta & LG confirm “next-gen XR device” partnership • UploadVR
David Heaney:
»
LG just officially announced an XR “strategic collaboration” with Meta.
Earlier today, Mark Zuckerberg met with LG CEO William Cho and the president of LG’s Home Entertainment division Park Hyoung-sei at LG’s headquarters in Seoul to finalize the details of the partnership. The meeting apparently included Zuckerberg demoing Quest 3 to Cho.
This is Zuckerberg’s first publicly-known trip to South Korea since 2014, when he visited Samsung to finalize the Gear VR smartphone-holder headset partnership.
William Cho, Mark Zuckerberg, and Park Hyoung-sei earlier today at LG headquarters in Seoul.
LG confirmed the talks included discussing “business strategies and considerations for next-gen XR device development”, giving the following statement:»
“LG envisions that by bringing together Meta’s platform with its own content/service capabilities from its TV business, a distinctive ecosystem can be forged in the XR domain, which is one of the company’s new business areas.
Moreover, the fusion of Meta’s diverse core technological elements with LG’s cutting-edge product and quality capabilities promises significant synergies in next-gen XR device development.”
«
«
One rather suspects that Zuck would have preferred to be visiting Samsung again, rather than smartphone-loser LG. But Samsung likely has its eyes on Google (whose blandishments Meta just rejected to cooperate on VR).
unique link to this extract
AI models make stuff up. How can hallucinations be controlled? • The Economist
»
Researchers at Google DeepMind found that telling an LLM to “take a deep breath and work on this problem step-by-step” reduced hallucinations and improved problem solving, especially of maths problems. One theory for why this works is that AI models learn patterns. By breaking a problem down into smaller ones, it is more likely that the model will be able to recognise and apply the right one. But, says Edoardo Ponti at the University of Edinburgh, such prompt engineering amounts to treating a symptom, rather than curing the disease.
Perhaps, then, the problem is that accuracy is too much to ask of llms alone. Instead, they should be part of a larger system—an engine, rather than the whole car. One solution is retrieval augmented generation (RAG), which splits the job of the ai model into two parts: retrieval and generation. Once a prompt is received, a retriever model bustles around an external source of information, like a newspaper archive, to extract relevant contextual information. This is fed to the generator model alongside the original prompt, prefaced with instructions not to rely on prior knowledge. The generator then acts like a normal LLM and answers. This reduces hallucinations by letting the LLM play to its strengths—summarising and paraphrasing rather than researching. Other external tools, from calculators to search engines, can also be bolted onto an LLM in this way, effectively building it a support system to enhance those skills it lacks.
Even with the best algorithmic and architectural antipsychotics available, however, LLMs still hallucinate. One leaderboard, run by Vectara, an American software company, tracks how often such errors arise. Its data shows that GPT-4 still hallucinates in 3% of its summaries, Claude 2 in 8.5% and Gemini Pro in 4.8%.
«
The RAG approach sounds like the adversarial system used for generating images such as thispersondoesnotexist, where one neural network generates and the other tries to find fault with it, feeding back between the two until the latter is satisfied.
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 