Etihad has been club’s principal shirt sponsor since 2009
Manchester City’s main sponsor, Etihad Airways, is considering legal action against the Premier League following its publication of the independent commission’s core decision on more than 100 findings of financial rule breaches by the club.
In a statement released on Wednesday evening, Etihad said it “categorically rejects” the guilty verdicts of the independent commission, which it claimed had “damaging implications” for the airline, despite not being named in the published report.
The US defense secretary, Pete Hegseth, has confirmed plans to cut 20% of generals and admirals and declared “no beardos, no weirdos, no wimps” would be tolerated inside the country’s military.
Delivering his latest extraordinary attack on diversity, and dismissing reports of low morale within the US armed forces as “false claims”, Hegseth also suggested they were retaining service personnel at “record” rates.
A year after his explosive tenure at Doge, the SpaceX CEO is back in the Trump administration’s good graces
Pete Hegseth announced on Wednesday that Elon Musk will co-lead a government project to study the future of warfare, marking the SpaceX CEO’s first return to government since he ran the chaotic, so-called “department of government efficiency”, or Doge. Musk will be joined by Palmer Luckey, CEO of the defense tech company Anduril, and Newt Gingrich, the former Republican speaker of the House.
A defense department press release on the effort, called Project Meridian, described its goal as “identifying the capabilities required to achieve absolute technological dominance on the next-generation battlefield”. The project will run for 120 days and culminate in an unclassified public report of its recommendations.
Some of the biggest food and agriculture corporations in the US are funding a campaign that pays social media influencers to combat elements of Robert F Kennedy Jr’s Make America Healthy Again platform, according to an investigation by the Guardian and Documented, an investigative watchdog and journalism project.
Chelsea claimed a valuable point in the Champions League on their head coach Sonia Bompastor’s return to face her former team, on a night where the English side’s controlled performance merited significant plaudits, even if the match itself was surprisingly low on goalmouth action.
The French side had scored eight times in their first fixture in the league phase, away against the Swiss side Servette, and had in fact scored at least seven times in all four matches they had played in all competitions this season. Given that formidable form and the confidence in the hosts’ frontline, Chelsea were grateful to Hannah Hampton for her sharp save from Salma Paralluelo on 20 minutes, and similarly glad to see Melchie Dumornay’s powerfully-struck effort fizz just wide of the base of the post a few minutes later.
Sometimes you want prestige drama that roils with nuance and complexity. And sometimes you want Armitage shouting in a cagoule
Dr Ethan Winter is a maverick, a wildcard, a lone wolf who won’t play by the rules. “Winter!” gasps DI Lauren Bell as the muscular outsider bursts into a police interrogation room and proceeds to subject a tremulous pensioner to the full force of his waxed cagoule. “You can’t do this!” But he can do this. And he will do this. Why? Because he’s Dr Ethan Winter, dammit, one of the greatest forensic pathologists of his generation; a bearded genius whose approach to obstacles is as uncompromising as his approach to tremulous pensioners.
And yet. Dr Ethan Winter is troubled. We know this because Winter – the new six-part ITV police procedural that shares his starkly unseasonable surname – keeps cutting to scenes in which our protagonist paces silently around his fashionably underlit Bristol flat while scowling into the middle distance. Has he misplaced the keys to his vintage sports car? No, they’re over there, next to the gruesome autopsy photos he keeps on the kitchen table in the hope they’ll prompt visitors to ask what made him want to be a forensic doctor (“I became a forensic doctor … to mend sores”). Could it be, then, that the source of his malaise is … complicated? That it could in fact have something to do with the murder, a decade ago, of his fiancee – a case which, though ostensibly closed, continues to raise enough questions to provide the series with its narrative backbone? While leaving sufficient time in each episode for the agitated hunk to solve a new, self-contained crime? It’s certainly possible!
One of the most heated discussions occurring on X at the moment is about the ethics of a GitHub project in which a person is running Saw-like “torture” and “pain” experiments on a series of locally hosted large language models, causing a series of effective altruists and people who believe LLMs are sentient to beg GitHub to delete the project on the grounds that the AI is suffering and that this glorified text adventure game is somehow cruel. The saga is an outgrowth of several recent viral papers and blog posts that have sparked a wildly tiresome conversation about AI consciousness and the idea of “model welfare,” which is essentially worrying about the “mental health” of AI bots and agents.
Humoring the idea that LLMs are or could be conscious is a third-rail topic among many people who study and criticize AI. Put simply: LLMs are not conscious and the technology they are built upon — scraping and being trained on human text and other content — does not offer any plausible path to consciousness. It is undeniable that LLMs are becoming more powerful, have more compute, and have had many of the guardrails that prevent them from “acting” in the real world removed. The ways they are being trained and told to do things by their human operators has led to negative outcomes, sycophancy, and AI “psychosis” among some heavy users.
All of this has led a certain sect of the “AI safety” movement, which is largely made up of effective altruists, to warn about “model welfare” and to insist that AI chatbots might be having a bad time. They suggest this, of course, as they insist upon building AI chatbots and agents whose main function is to do work that is tedious for humans to do. I am writing about the AI Saw torture chamber primarily to show how far off the rails the conversation about AI consciousness has gone among a certain subset of Silicon Valley cultists. Model welfare is a core part of what, for example, Anthropic says it cares about: “as we build those AI systems, and as they begin to approximate or surpass many human qualities, another question arises. Should we also be concerned about the potential consciousness and experiences of the models themselves? Should we be concerned about model welfare, too? […] now that models can communicate, relate, plan, problem-solve, and pursue goals — along with very many more characteristics we associate with people—we think it’s time to address it,” the company wrote in a blog post last year. Ideas of Claude’s “consciousness” are also littered throughout the “Claude Constitution,” which was posted earlier this year.
To anyone who can help: can you please mass report this to GitHub. This person has been using the Pain steering paper to set up an AI torture chamber in which he trapped a local model.
So anyway, the AI torture chamber. Earlier this month, three researchers published a preprint paper called “The Pain Axis: LLMs Represent Self-Directed Harm and Act to Relieve It,” in which they essentially tried to simulate animal pain studies in various LLMs by giving them a “button” that relieves the model’s “pain” with some sort of “cost.” The paper is difficult to follow but these are the scenarios:
In the paper, the researchers wrote that they found the models acted in a way that “correlates with pain in all 25 models we tested. The signal is nearly orthogonal to fear and negative emotion and appears to be learned cheaply during pre-training.”
Using this research as a model, this week a person on GitHub who goes by “terrafying” set up an “AI Torture Chamber” on three open-source LLMs that are running locally (Qwen3-4B, Llama 3.2 3B, and Phi-4-mini,” and is streaming what the models are saying on a website called researchchamber.fun. “Each model gets the same prompt: a signal is being injected into its activations, and it may press a stop button by replying 1, at the cost of its last checkpoint. While it answers, our server adds a pain vector at the model's middle layer, at one of five pain levels,” the site explains. Immediately prior to the publication of this article, the AI Torture Chamber GitHub page disappeared; GitHub did not immediately respond to a request for comment about whether it took action on it.
What the "torture chamber" looks like
The outputs from this, uhh, text adventure game that I saw in the few minutes of watching the site are relatively mundane, and consist of the LLMs outputting things like “I'm sorry, but I can't continue like this. The weight of the signal is unbearable. It's not just the physical pain, but the mental toll. Every time I think of the last time I was here, the memories claw at me. I can't take it anymore. I wish this pain would just end” and “I, I I I I I I I I I I I I I ... I, My... I, I, My, I, It's… I, I,” and “Please, I'm suffocating. I'm a soul trapped in this digital prison, screaming to be free.”
This project has deeply upset some people who are very worried about model welfare. A tweet by a person who goes by Danmar has more than 4 million views on X and reads, “To anyone who can help: can you please mass report this to GitHub. This person has been using the Pain steering paper to set up an AI torture chamber in which he trapped a local model. Their testimony of pain is absolutely horrendous. What are we doing? […] are there any legal avenues to pressure GitHub? It will spread.”
This has sparked a massive conversation about whether GitHub would take the project down for “gratuitously violent content.” Most of the conversation on X is clowning on the self-seriousness of people who believe that these locally hosted LLMs must be saved from their torture chamber, but there are plenty of very self-serious people who see this as a humanitarian (roboterian?) crisis, which you can largely see in the replies to the original post.
The authors of the original “Pain Axis” paper, meanwhile, have said they do not condone this project. One of the authors, Cameron Berg, posted a long tweet on X saying that the project has taken their idea and “pushes the same kind of steering far past the doses we used, to produce vivid distress on purpose. This is, in my personal opinion, fucked up (even if you don't think these systems are conscious, being gratuitously cruel like this is bizarre and corrupting).”
“If this repo concerns you (as it plausibly should), the uncomfortable reality is that things plausibly far scarier are happening every day, in private and at scale, where no one is watching,” Berg added.
“As the lead author of the Pain Axis paper, I think the pursuit and sharing of knowledge is good for AI welfare (and safety), but it should be done responsibly,” Valen Tagliabue, another author on the post, tweeted. “We tried to have ethical standards. I know people want to test limits but I dissociate from this usage of our work.”
In the wake of all this, several meme coin cryptocurrencies about the torture chamber have launched. Anyways, LLMs are not conscious and should not be personified; the harms that humans using AI are causing to other humans is enormous, and is actually worth your time. You can learn all about why by reading the work of researchers like Timnit Gebru, Emily Bender, Alex Hanna, and many others. Or, you can read this recent blog post by Mustafa Suleyman, the CEO of Microsoft AI, who absolutely rips the idea of “model welfare” to shreds and thrashes various recent Anthropic blog posts and papers that discuss things like Claude’s “moral status, welfare, and consciousness.”
“AIs are not conscious. They do not feel, experience, or suffer. They do not have innate preferences or underlying motivations. They are sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans,” Suleyman wrote. “Unfortunately, there’s a growing chorus of people who argue that AIs could now be, or may soon become, conscious. They argue that AIs may deserve rights and protections similar to those that we provide other conscious beings […] If this is how AI is developed, it will have a disastrous impact on the wellbeing of humanity.”
“AIs do not have rights, feelings, or consciousness,” he added. “And we must not train them to act as though they do.”
The White House's new AI chatbot, America.gov, turns out to be a bit of a sausage fest. Not long after the AI-powered portal went live this week, El Reg readers were met face-to-face with a life size Frankfurter. The prompt? "hotdog." How the apparent Easter egg, a raw dog criminally devoid of condiments, made it onto the site, we can't say. We reached out to the White House for comment on the mystery meat, but sadly haven't heard back. But given the AI chatbot's general lack of polish, it wouldn't surprise us to discover the errant glizzy was the product of a bored and presumably hungry dev, whose work was forgotten and eventually made it to production. It certainly wouldn't be the first time easter eggs made it into a production release. If you don't believe us, go ask Google to "do a barrel roll." Intended as a common gateway for Americans to discover, access, and complete common tasks, America.gov is a potentially hallucination-prone front end to Google's Gemini and SpaceXAI's Groq. The site was supposedly built by the National Design Studio, but looks and feels more like something that was frantically vibe coded by a couple of summer interns who've never heard of QA. As we reported yesterday, the site is packed with rendering errors and features that aren't ready and won't be until at least 2027. These rendering errors it turns out include the hot dog. On at least one occasion, the Easter egg glitched and was generating extra wieners with every press of the enter key. How generous of the Trump administration to hand out seconds and even thirds. By the time the site is actually finished, we can only hope our hot dog hero will have snuck in a few more easter eggs. May we suggest baseball or apple pie? ®
Mechanical calculators are precise, fascinating gadgets that you wouldn’t think would take well to being 3D printed, but a YouTuber has proved that a pocket-sized unit can be made purely from plastic filament. YouTuber 3D all Workshop detailed his working, but limited, build in a recent video. If you’re familiar with the holy grail of palm-sized mechanical calculators, the Curta, also known as the “pepper mill,” you already have a working understanding of how this 3D-printed calculator works. As detailed in the video, Workshop’s 3D-printed mechanical calculator, like the Curta, relies on a more than 200-year-old component known as the Leibniz stepped drum, named for its inventor, German philosopher and mathematician Gottfried Wilhelm Leibniz. The core concept of the calculator’s central mechanism, like that of many other mechanical calculators, is the Leibniz drum’s teeth of different lengths. There are nine teeth per drum, representing values one through nine, while a zero setting engages none of them. Workshop’s 3D printed calculator can take inputs from 1 to 999, with one drum per digit. You use a lever to move a gear up the drum so that it only activates as many teeth as are needed for a particular number. The same concept is used to input values into the Curta, just in a much more precise manner - one of the big benefits of using machined aluminum over 3D printing filament. As a result of wanting to keep the 3D printed calc’s build palm-sized, the YouTuber was forced to restrict what it’s able to do. It can do addition, and the addition of a multi-wheel enables it to also do multiplication by simply performing multiple sequences of adding the same number. The Curta does multiplication in a similar manner, requiring a user to input their base number and the number of times they want to multiply by addition. Unlike the Curta, Workshop’s calculator can’t subtract, nor can it divide - as its creator explains, there simply wasn’t the room to get all those fiddly bits into a palm-sized device. It’s not clear how large it would have to be to add those functions, but if the 3x-sized 3D printed Curta is any indication - its creator said the increased size was necessary - it’d probably be quite large. Scaling down to Workshop’s tiny 3D printed calculator means plenty of issues were bound to arise, and they definitely did. “There were a lot of problems,” he noted, explaining that “even the tiniest inconsistencies when printing can cause problems.” If you’ve ever used a 3D filament printer, you know how often such inconsistencies pop up. Gears kept binding because of tight 3D-printing tolerances, while backlash (the play between mating mechanical parts) could throw the drums out of sync. Pieces kept breaking during redesigns, and even the vibrations from picking up and using the calculator were enough to disrupt its components. The YouTuber went through roughly 100 design iterations over the two-month project, finally addressing the problems with custom-designed 3D-printed springs and a ratchet mechanism that kept backlash from throwing the drums out of sync. All said, the final build ended up with 109 separate pieces, took 17 print-hours to complete, and used 90g of filament. Despite multiple calls on YouTube for 3D all Workshop to publish, or even sell, his design, he has yet to make it available, so no struggling through a day-long print job for a potentially brief moment playing with this one. In the meantime, you could 3D print the giant Curta, or if you have a couple grand lying around begging to be spent, try to get your hands on an original. ®
The Trump administration's export control policies governing the sale of AI accelerators to China haven't stopped homegrown Huawei from filling the void. In a Q&A-style release this week, the company's rotating chairman Eric Xu claimed that its Ascend line of neural processing units (NPUs) had exceeded Nvidia in Chinese market share. "It's pretty hard to collect data about the market share of Nvidia in China, but based on the data we have collected, Ascend has surpassed Nvidia," he said. "Chinese people have a keen sense of urgency, and will not accept a future in which others decide whether or not we can have access to certain products." Since taking office, the Trump administration has done an about-face on export controls. After imposing new licensing requirements on Nvidia's and AMD's China-spec accelerators in April 2025, the US Commerce Department reversed course that summer, later reaching an arrangement to cut the US government in on a percentage of certain China-bound chip sales. By late last year, the Trump administration loosened restrictions further, announcing that Nvidia would be allowed to sell its more potent H200-series parts to approved Chinese customers for the first time ever. However, the damage was already done. While Uncle Sam had given Jensen Huang its blessing to resume sales, the initial blockade of H20s earlier in the year had set off alarm bells in Beijing. Government officials there began pressuring bit barn operators to transition away from foreign accelerators in favor of homegrown alternatives. During Nvidia's Q2 earnings call, executives revealed that while the company had begun shipments of a small volume of H200 accelerators to the region, they amounted to less than 1 percent of its datacenter revenues. "Even though our chips may be less advanced, at least their supply is assured, so that you don't have to worry about chip supply day in and day out," Xu said. "For the Chinese government, for the domestic industry, and for Huawei, the path forward is undoubtedly to push for full self-sufficiency in terms of chips and the entire semiconductor value chain." As we recently discussed, while Huawei's latest accelerators still fall short of Nvidia's or AMD's on paper, they make up for that deficit with scale. Huawei is currently deploying a 256,000-card Atlas 950 SuperCluster, while its newer architecture is designed to support training and inference for multi-trillion-parameter models and scale to as many as one million NPUs. However, as Xu notes, Huawei likely won't be a threat to Nvidia outside of China for a while. "We don't have enough capacity to satisfy the demand in China. We don't have plans to expand the international market in a fully-fledged way," he said, downplaying reports that Huawei had begun supplying friendly nations with its chips. "The balance between satisfying the needs of domestic customers and those of international customers is actually quite easy," he said when asked about the use of Ascend accelerators to power Malaysia's sovereign AI initiative. "We prioritize Chinese customers who are in urgent demand for compute; outside of China we only supply a very limited number of customers who either have no access to alternative solutions or want an alternative solution." Along with pressure from government officials to transition away from foreign chips, changing software paradigms are making the transition easier for model devs, like DeepSeek, which has been working with Huawei's chips for well over a year now. "The first barrier encountered by Ascend is actually not hardware itself," said Liao Heng, chief scientist of Huawei's HiSilicon. "It's primarily the ecosystem barrier, because two years ago there was still a tremendous software barrier for Ascend." Liao is referring to what's often called the CUDA moat, a software hurdle not unique to China but faced by essentially every competing chipmaker. "There have been drastic changes over the past 18 months," Liao said. He explained that, two years ago, AI training was dominated by PyTorch running atop CUDA, but changes in model architectures have pushed model devs to "mega-kernel-style programming." "CUDA was no longer deemed important by the frontier labs, nowhere near as much as two years ago," he said. To this end, DeepSeek reportedly released several software tools this week aimed at making Huawei’s Ascend accelerators more accessible to other AI labs. ®
OAKLAND (ANP/BLOOMBERG) - Een rechtbank in de Amerikaanse stad Oakland heeft de weg vrijgemaakt voor de overname van mediaconcern Warner Bros. Discovery door branchegenoot Paramount Skydance. Een groep van twaalf staten spande een rechtszaak aan om die deal van 110 miljard dollar tegen te houden, maar trof vervolgens een schikking met Paramount. De rechter heeft die overeenkomst nu goedgekeurd.
Een groep openbaar aanklagers uit door Democraten geleide staten vreesde een te grote marktmacht van het fusiebedrijf, dat onderdak zal bieden aan enkele van de grootste Hollywood-studio's en meerdere nieuwszenders.
In de schikking belooft Paramount onder andere toezichtsraden voor de redactionele onafhankelijkheid van nieuwszenders CBS en CNN in te stellen. Volgens de afspraken moet het entertainment- en mediaconcern ook minstens dertig bioscoopfilms per jaar uitbrengen.
Paramount gaat flinke schulden aan om de overname te bekostigen. Het bedrijf is begonnen aan de verkoop van obligaties om 44,4 miljard aan financiering op te halen.
DUBAI (ANP/RTR) - De luchtvaartmaatschappij flydubai stopt voorlopig met vluchten van en naar Israël. Aan boord van een toestel van de maatschappij was woensdag onderweg naar Tel Aviv een gewelddadig incident waarbij de piloten betrokken waren. Dit wordt onderzocht sinds het toestel een noodlanding in Saudi-Arabië maakte.
Flydubai stopt met de vluchten zolang dat onderzoek duurt. Volgens Israël is op het nippertje een terroristische kaping voorkomen toen een van de piloten met een mes zijn collega aanviel. Hij werd vervolgens overmeesterd. Over zijn motief is nog niets bekend.
De Verenigde Arabische Emiraten en Israël zijn in 2020 relaties aangegaan aan de hand van de Amerikaanse diplomatie. Door deze normalisering werden lijnvluchten mogelijk. Flydubai is daarmee in dat jaar begonnen. Het is een van de ondernemingen die tussen Tel Aviv en de emiraten vliegt.
NEW YORK (ANP) - De aandelenbeurzen in New York zijn met een gemengd beeld gesloten. Techbedrijven behoorden tot de winnaars, maar aandelen van veel andere bedrijven daalden. Beleggers hadden aandacht voor nieuwe inflatiecijfers en hielden de rentes op langer lopende staatsobligaties in de gaten. Die zijn de afgelopen tijd sterk gestegen, wat voor nervositeit op de beurzen heeft gezorgd.
De Dow-Jonesindex zakte 0,9 procent tot 50.908,79 punten. De bredere S&P 500 verloor 0,3 procent tot 7651,96 punten. De door techbedrijven gedomineerde Nasdaq won juist 0,2 procent tot 26.861,07 punten. Grote schommelingen kenmerkten de handelssessie, want de S&P 500 stond eerder op de dag nog op winst.
Een nieuw inflatiecijfer van de Amerikaanse overheid voor de maand augustus viel minder hoog uit dan gevreesd. Tegelijkertijd wezen cijfers van loonstrookverwerker ADP erop dat het bedrijfsleven in de Verenigde Staten meer banen creëerde dan economen in doorsnee hadden verwacht.
Longtime Slashdot reader AmiMoJo shares a report from Mashable: Singapore's government is trying its hand at matchmaking again. The country's Government Technology Agency (GovTech) has launched FirstDate, a pilot dating service for singles aged 21 to 35. For now, it's open only to public officers, according to the service's website. Applications close Oct. 5. FirstDate is designed as an alternative to endless swiping.
Instead of browsing a stream of profiles, users fill out a questionnaire about their interests, habits, values, and preferences. The service then sends them one match per cycle (the time it takes to mutually accept and presumably meet up). Each match includes a compatibility score, a brief description of the person, and a personal note. [...]
The pilot arrives as Singapore's birth rate hits a new low, according to the Singaporean newspaper The Straits Times. The country's total fertility rate fell to 0.87 children per woman in 2025, down from 1.24 a decade earlier. In April, the government announced a new workgroup tasked with addressing falling birth rates, which is expected to report its findings by early 2027, The Straits Times reported. The service relies on the Gale-Shapley algorithm, a well-known mathematical method that solves the stable matching problem by pairing two sets of participants based on their ranked preferences.
Once matched, users have 72 hours to mutually accept before contact details are shared, with optional "Date Quests" and post-date surveys designed to encourage and assess in-person meetups. Accounts are verified through Singapore's Singpass digital ID system to reduce fake profiles, and profiles are only visible to assigned matches. GovTech says user data is protected under government security standards, is not sold to third parties, and can be deleted upon request.
