An anonymous reader quotes a report from the Associated Press: Japan's nuclear watchdog said Wednesday it is scrapping the safety screening for two reactors at the Hamaoka nuclear power plant in central Japan, after its operator was found to have fabricated data about earthquake risks. It was a setback to Japan's attempts to accelerate nuclear reactor restarts. Less than a quarter of commercial nuclear reactors are operational in the wake of the 2011 Fukushima Daiichi meltdowns, but rising energy costs and pressure to reduce carbon emissions have pushed the government to prioritize nuclear power. Chubu Electric Power Co. had applied for safety screening to resume operations at the No. 3 and 4 reactors at the Hamaoka plant in 2014 and 2015. Two other reactors at the plant are being decommissioned, and a fifth is idle. The plant, about 200 kilometers (125 miles) west of Tokyo, is located on a coastal area known for potential risks from so-called Nankai Trough megaquakes. The Nuclear Regulation Authority said it started an internal investigation last February, after receiving a tip from a whistleblower that the utility had for years provided fabricated data that underestimated potential seismic risks. The regulator suspended the screening for the reactors after it confirmed the falsification and the utility acknowledged the fabrication in mid-December, said Shinsuke Yamanaka, the watchdog's chair. The NRA is also considering inspecting the utility headquarters. [...] The scandal surfaced Monday when Chubu Electric President Kingo Hayashi acknowledged that workers at the utility used inappropriate seismic data with an alleged intention to underestimate seismic risks. He apologized and pledged to establish an independent panel for investigation. The screening, including data that had been approved earlier, would have to start from scratch or possibly be rejected entirely, Yamanaka said. The NRA will decide on the case next week, without waiting for the utility's probe results, he said. "Ensuring safety is the first and foremost responsibility for nuclear plant operators," Yamanaka said. "It is outrageous and it's a serious challenge to safety regulation."

Read more of this story at Slashdot.

Samsung Electronics and SK hynix are projected to raise server memory prices by up to 70% in early 2026, according to Korea Economic Daily. "Combined with 50 percent increases in 2025, this could nearly double prices by mid-2026," reports the Register. From the report: The two Korean giants, alongside US-based Micron, dominate global memory production. All three are reallocating advanced manufacturing capacity to high-margin server DRAM and HBM chips for AI infrastructure, squeezing supply for PCs and smartphones. Financial analysts have raised their earnings forecasts for the firms in response, as they look to benefit from the AI infrastructure boom that is driving up prices for everyone else. Taiwan-based market watcher TrendForce reports that conventional DRAM prices already jumped 55-60 percent in a single quarter. Yet despite the focus on server chips, supply of these components continues to be strained too, with supplier inventories falling and shipment growth reliant on wafer output increases, according to TrendForce. As a result, it forecasts that server DRAM prices will jump by more than 60 percent in the first quarter of 2026. Prior to Christmas, analyst IDC noted the "unprecedented" memory chip shortage and warned this would have knock-on effects for both hardware makers and end users that may persist well into 2027.

Read more of this story at Slashdot.

Google and Character.AI have agreed to settle multiple lawsuits from families alleging the chatbot encouraged self-harm and suicide among teens. "The settlements would mark the first resolutions in the wave of lawsuits against tech companies whose AI chatbots encouraged teens to hurt or kill themselves," notes Axios. From the report: Families allege that Character.AI's chatbot encouraged their children to cut their arms, suggested murdering their parents, wrote sexually explicit messages and did not discourage suicide, per lawsuits and congressional testimony. "Parties have agreed to a mediated settlement in principle to resolve all claims between them in the above-referenced matter," one document filed in U.S. District Court for the Middle District of Florida reads. The documents do not contain any specific monetary amounts for the settlements. Pricy settlements could deter companies from continuing to offer chatbot products to kids. But without new laws on the books, don't expect major changes across the industry. Last October, Character.AI said it would bar people under 18 from using its chatbots, in a sweeping move to address concerns over child safety.

Read more of this story at Slashdot.

FEX, the open-source emulator for running x86 and x86_64 binaries on AArch64 (ARM64) Linux and that is sponsored by Valve and to be used by the Steam Frame, is out with a new monthly feature release...

OpenAI has unveiled ChatGPT Health, a sandboxed health-focused mode that lets users connect medical records and wellness apps for more personalized guidance. The company makes sure to note that ChatGPT Health is "not intended for diagnosis or treatment." The Verge reports: The company is encouraging users to connect their personal medical records and wellness apps, such as Apple Health, Peloton, MyFitnessPal,Weight Watchers, and Function, "to get more personalized, grounded responses to their questions." It suggests connecting medical records so that ChatGPT can analyze lab results, visit summaries, and clinical history; MyFitnessPal and Weight Watchers for food guidance; Apple Health for health and fitness data, including movement, sleep, and activity patterns"; and Function for insights into lab tests. On the medical records front, OpenAI says it's partnered with b.well, which will provide back-end integration for users to upload their medical records, since the company works with about 2.2 million providers. For now, ChatGPT Health requires users to sign up for a waitlist to request access, as it's starting with a beta group of early users, but the product will roll out gradually to all users regardless of subscription tier. [...] In a blog post, OpenAI wrote that based on its "de-identified analysis of conversations," more than 230 million people around the world already ask ChatGPT questions related to health and wellness each week. OpenAI also said that over the past two years, it's worked with more than 260 physicians to provide feedback on model outputs more than 600,000 times over 30 areas of focus, to help shape the product's responses. "ChatGPT can help you understand recent test results, prepare for appointments with your doctor, get advice on how to approach your diet and workout routine, or understand the tradeoffs of different insurance options based on your healthcare patterns," OpenAI claims in the blog post.

Read more of this story at Slashdot.

An anonymous reader quotes a report from TechCrunch: Senator Steve Padilla (D-CA) introduced a bill [dubbed SB 867] on Monday that would place a four-year ban on the sale and manufacture of toys with AI chatbot capabilities for kids under 18. The goal is to give safety regulators time to develop regulations to protect children from "dangerous AI interactions." "Chatbots and other AI tools may become integral parts of our lives in the future, but the dangers they pose now require us to take bold action to protect our children," Senator Padilla said in a statement. "Our safety regulations around this kind of technology are in their infancy and will need to grow as exponentially as the capabilities of this technology do. Pausing the sale of these chatbot-integrated toys allows us time to craft the appropriate safety guidelines and framework for these toys to follow." [...] "Our children cannot be used as lab rats for Big Tech to experiment on," Padilla said.

Read more of this story at Slashdot.

According to the Wall Street Journal (paywalled), Goldman Sachs is transferring Apple Card and Apple Savings to JPMorgan Chase. "It was clear in 2023 that Goldman Sachs would exit the consumer credit game, abandoning its Apple Card partnership with it," reports AppleInsider. "However, it has taken 26 months to reach a point where it can finally hand over issuing control to another bank." From the report: Goldman Sachs is reportedly expected to hand over the $20 billion of outstanding balances at a $1 billion discount. Such discounts are rare, and allegedly reflect the higher-than-average delinquency rate found with Apple Card holders. JPMorgan will have to issue new Apple Cards to existing users, but it may be some time before that is done. A new Apple Savings will be opened by JPMorgan as well, but users will be given the option to move or stay.

Read more of this story at Slashdot.

Bose is end-of-lifing its SoundTouch smart speakers but softened the blow by open-sourcing the SoundTouch API and preserving limited local features, AirPlay, and Spotify Connect. Ars Technica reports: In October, Bose announced that its SoundTouch Wi-Fi speakers and soundbars would become dumb speakers on February 18. At the time, Bose said that the speakers would only work if a device was connected via AUX, HDMI, or Bluetooth (which has higher latency than Wi-Fi). After that date, the speakers would stop receiving security and software updates and lose cloud connectivity and their companion app, the Framingham, Massachusetts-based company said. Without the app, users would no longer be able to integrate the device with music services, such as Spotify, have multiple SoundTouch devices play the same audio simultaneously, or use or edit saved presets. The announcement frustrated some of Bose's long-time customers, some of whom own multiple SoundTouch devices that still function properly. Many questioned companies' increasingly common practice of bricking expensive products to focus on new devices or to minimize costs, or because they've gone through acquisitions or bankruptcy. SoundTouch speakers released in 2013 and 2015 with prices ranging from $399 to $1,500. Today, Bose had better news. In an email to customers, Bose announced that AirPlay and Spotify Connect will still work with SoundTouch speakers after EoL, expanding the wireless capabilities that people will still be able to access. Additionally, SoundTouch devices that support AirPlay 2 can play the same audio simultaneously. The SoundTouch app will also live on, albeit stripped of some functionality. "On May 6, 2026, the app will update to a version that supports the functions that can operate locally without the cloud. No action will be required on your part. Opening the app will apply the update automatically," Bose said. Bose also provided instructions (PDF) for a workaround for saving presets that uses the favorites options in music service apps.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Reuters: Warner Bros Discovery's board unanimously turned down Paramount Skydance's latest attempt to acquire the studio, saying its revised $108.4 billion hostile bid amounted to a risky leveraged buyout that investors should reject. In a letter to shareholders on Wednesday, Warner Bros' board said Paramount's offer hinges on "an extraordinary amount of debt financing" that heightens the risk of closing. It reaffirmed its commitment to streaming giant Netflix's $82.7 billion deal for the film and television studio and other assets. Their assessment comes even after Paramount, which has a market value of around $14 billion, proposed to use $40 billion in equity personally guaranteed by Oracle billionaire co-founder Larry Ellison -- father of Paramount CEO David Ellison -- and $54 billion in debt to finance the deal. The decision keeps Warner Bros on track for its deal with Netflix, even after Paramount amended its bid on December 22 to address the earlier concerns about the lack of a personal guarantee from Larry Ellison. Netflix co-CEOs Ted Sarandos and Greg Peters welcomed Warner Bros' decision on Wednesday, saying it recognizes the streaming giant's deal "as the superior proposal that will deliver the greatest value to its stockholders, as well as consumers, creators and the broader entertainment industry."

Read more of this story at Slashdot.

The humble power bank has transformed from a simple pocket-sized battery into a feature-laden gadget that now sometimes includes screensavers, Bluetooth connectivity and built-in Wi-Fi hotspots. The Verge's Thomas Ricker highlighted the $270 EcoFlow Rapid Pro X Power Bank 27k at CES 2026 as a prime offender -- a device he declared "too expensive, too big, too slow, and too heavy." Its giant display takes 30 seconds to wake from sleep, plays swirly graphics and blinking eyeballs, and requires a screensaver while slowly draining the battery it's meant to preserve. The feature creep is industry-wide. Anker no longer lists a display-less model in its 20,000mAh range, and both companies sell proprietary desk chargers. Basic alternatives exist -- Anker's PowerCore 10k runs $26 -- but they're becoming harder to find.

Read more of this story at Slashdot.

For those using a DisplayPort to HDMI dongle currently with the AMDGPU Linux kernel graphics driver may find some higher resolutions / modes unavailable. Fortunately, a fix is on the way for dealing with this situation due to an oversight in the kernel driver...

Hier, le 6 janvier 2026, nous vous présentions pour l'ouverture du CES une GeForce RTX 5090 particulière à plus d'un titre : la GIGABYTE GeForce RTX 5090 AORUS INFINITY. Son design "arrondi aux deux extrémités" n'est pas commun du tout, mais d'un point de vue technique l'information la plus importan...

An anonymous reader shares a report: Ever since the federal government began issuing the Dietary Guidelines in 1980, it has told Americans to limit themselves to one or two standard alcoholic drinks a day. Over time, the official advice morphed to no more than two drinks a day for men, and no more than one for women. No longer [non-paywalled source]. The updated guidelines issued on Wednesday say instead that people should consume less alcohol "for better overall health" and "limit alcohol beverages," but they do not recommend clear limits. The guidelines also no longer warn that alcohol may heighten the risk of breast cancer and other malignancies. It is the first time in decades that the government has omitted the daily caps on drinking that define moderate consumption -- standards that are used as benchmarks in clinical studies, to steer medical advice, and to distinguish moderate from heavy drinking, which is unquestionably harmful. The new guidance advises Americans who are pregnant, struggle with alcohol use disorder or take medications that interact with alcohol to avoid drinking altogether. The guidelines also warn people with alcoholism in the family to "be mindful of alcohol consumption and associated addictive behaviors." They do not, however, distinguish between men and women, who metabolize alcohol differently, nor do they caution against underage drinking. The guidelines also no longer include a warning that was in the last set issued in 2020: that even moderate drinking may increase the risk of cancer and some forms of cardiovascular disease, as well as the overall risk of dying.

Read more of this story at Slashdot.

Samsung Electronics has once again sidelined Ballie, a long-anticipated robot that was first announced six years ago but never released. Bloomberg News: The device -- designed to roll and roam throughout the home -- is completely absent from this week's CES, the biggest electronics trade show. And though Samsung said last year that Ballie was nearly ready for a retail release, the product is now unlikely to resurface soon. In an emailed statement, Samsung referred to Ballie as an "active innovation platform" within the company, rather than a forthcoming consumer device. "After multiple years of real-world testing, it continues to inform how Samsung designs spatially aware, context-driven experiences, particularly in areas like smart home intelligence, ambient AI and privacy-by-design," a Samsung spokesperson said in the statement.

Read more of this story at Slashdot.

Several years after Samsung introduced the Frame TV in 2017 -- a television designed to display fine art and resemble a framed painting when switched off -- competitors are finally catching up in meaningful numbers. Amazon announced the Ember Artline TV at CES 2026 this week, a $899 model that can display one of 2,000 works of art for free and includes an Alexa AI tool to recommend pieces suited to your room. Hisense unveiled its CanvasTV late last year, TCL has the NXTvision model, and LG has announced the Gallery TV for later this year. The surge in art-focused televisions comes down to two factors: smaller living spaces in cities where younger buyers lack dedicated rooms for large screens, and advances in matte screen technology that enable displays to absorb light like a canvas rather than reflect it like a window. Local dimming and improved backlighting processing allow these newer models to maintain their slim profiles for flush wall-mounting while delivering more realistic art reproduction than earlier edge-lit designs.

Read more of this story at Slashdot.

While for years open-source firmware enthusiasts have been after an open-source Firmware Support Package "FSP" for Intel CPUs and back during Raja Koduri's tenure at Intel it sounded like it might happen, it has yet to happen. But at least with the forthcoming Intel Core Ultra Series 3 "Panther Lake" there are some FSP improvements...

Le tarif du Minisforum AI X1 Pro-470 n’est pas encore connu, pas plus que sa date de sortie effective. Une seule chose est quasi sure, il ne sera pas donné. Là, je ne prends pas trop de risques, entre la puce Ryzen AI 9 HX 470 et le reste des composants, même un gnou peu informé sur l’évolution des prix, sait que le marché n’est pas facile en ce moment.

L’engin promet beaucoup de possibilités techniques avec des évolutions assez intéressantes qui reflètent bien le marché actuel. La première est une grosse emphase sur l’IA. Minisforum met en avant le NPU de la puce en confondant quelque peu ses capacités internes (55 TOPS) avec celles plus globales de la puce entière (86 TOPS). Cela ferait du MiniPC une station IA complète, adoubée par Microsoft dans sa gamme Copilot PC. De quoi piloter des outils IA en local sous Windows d’après les deux marques.

On n’a pas le détail de tous les composants mais le Minisforum AI X1 Pro-470 promet jusqu’à 128 Go de mémoire vive DDR5 et embarque trois emplacements M.2 NVMe PCIe 4.0 pour proposer jusqu’à 12 To de stockage. Un ensemble de composants qui rendra le prix de la machine elle-même totalement accessoire.

D’autres éléments sont listés comme la présence d’un lecteur d’empreintes digitales, d’un modem Wi-Fi7 et Bluetooth 5.4, d’une connectique Ethernet 2.5 Gigabit et de la possibilité de piloter quatre écrans en UltraHD. Un port OCuLink est également présent pour seconder la minimachine avec un dock graphique complet. 

Deux détails originaux font écho à un autre constructeur et pourraient devenir une tendance assez forte dans le futur. À l’instar de Beelink qui a basculé ses dernières machines de la même manière depuis le début de l’année 2025. Le Minisforum AI X1 Pro-470 embarque désormais son alimentation en interne. On pourra donc le connecter directement au secteur. De la même façon, le constructeur a décidé de proposer un système audio en interne.

La coque semble être en aluminium sablé avec, en façade, deux ports USB 3.2 Type-A,  un USB type-C et un jack audio combo 3.5 mm. Deux micros sont visibles pour écouter vos ordres. Un bouton classique est présent à gauche, séparé du lecteur d’empreintes et collé sur le haut du châssis. Il est doublé d’un bouton Copilot à droite. Une idée toujours aussi fabuleuse, probablement générée par une IA Microsoft, qui poussera encore sans doute à éteindre ou mettre en veille de nombreux PC. Coller un bouton sur une machine qui obligera à lâcher sa souris pour cliquer dessus alors qu’on a une icône sur son bureau qui fera la même tâche, cela résume assez bien l’efficacité de ce concept.

Les capacités de calcul IA de la nouvelle génération AMD ne sont pas très éloignées de celles de la génération précédente d’après l’exemple donné sous DeepSeek. Si vous avez déjà un MiniPC sous Ryzen HX 370, le passage au HX 470 ne semble pas forcément être une bonne idée.

Un Minisforum AI X1 Pro-470 annoncé très en avance ?

Comme d’habitude, les annonces des constructeurs de MiniPC au CES doivent être prises avec le recul de rigueur. Il s’agit souvent de semer de petites graines d’attente dans l’esprit du public. Chacun publie des images 3D d’une machine en espérant être le premier à réellement la dégainer. Il est fort possible que l’engin ne soit pas disponible avant un bon moment. L’important c’est que le client du futur pense à comparer toutes les solutions de cette gamme à celle-ci au cours des mois prochains.

On ne connait pas les moyens mis par AMD pour déployer les puces Gorgon Point. Mais au vu de l’historique Strix Point, il est possible que cela soit assez compliqué de monopoliser les usines de TSMC en ce moment. Alors est-ce qu’AMD va privilégier les petits acteurs comme Minisforum et consorts pour ce lancement ou focaliser les livraisons possibles à ses clients historiques les plus importants comme HP, Lenovo et autres marques dans le Top10 des constructeurs de portables ? Je suppose que vous avez déjà la réponse à cette question.

Il faudra certainement attendre quelque temps avant que les machines soient disponibles, chez Minisforum comme chez les concurrents.

Source Minisforum

Minisforum AI X1 Pro-470 : le premier MiniPC Ryzen AI 9 HX 470 © MiniMachines.net. 2025

Il a rapidement fallu se rendre à l'évidence, puisqu'AMD a levé très tôt le mardi 6 janvier son NDA sur ses nouveautés du CES 2026 : pas de Ryzen 9 9950X3D2 en vue, alors qu'il était attendu de pied ferme par ceux qui espéraient voir en action une double 3D V-Cache pour la première fois sur un CPU g...

Climate-heating emissions from aviation could be slashed in half -- without reducing passenger journeys -- by getting rid of premium seats, ensuring flights are near full and using the most efficient aircraft, according to analysis. The Guardian: These efficiency measures could be far more effective in tackling the fast-growing carbon footprint of flying than pledges to use "sustainable" fuels or controversial carbon offsets, the researchers said. They believe their study, which analysed more than 27m commercial flights out of approximately 35m in 2023, is the first to assess the variation in operational efficiency of flights across the globe. The study, led by Prof Stefan Gossling at Sweden's Linnaeus University, examined flights between 26,000 city pairs carrying 3.5 billion passengers across 6.8 trillion kilometers. First and business class passengers are responsible for more than three times the emissions of economy travelers, and up to 13 times more in the most spacious premium cabins. The average seat occupancy across all flights in 2023 was almost 80%. US airports accounted for a quarter of all aviation emissions and ran 14% more polluting than the global average. Atlanta and New York ranked among the least efficient airports overall, nearly 50% worse than top performers like Abu Dhabi and Madrid.

Read more of this story at Slashdot.

Microsoft has canceled plans to impose a daily limit of 2,000 external recipients on Exchange Online bulk email senders. From a report: The change was announced in April 2024, when Microsoft said that it would add new External Recipient Rate (ERR) limits starting January 2025 to fight spam, with plans to begin enforcing the limit on cloud-hosted mailboxes of existing tenants between July and December 2025. As explained last year, this new Mailbox External Recipient Rate Limit was designed to prevent Microsoft 365 customers from abusing Exchange Online resources and to restrict unfair usage. However, on Tuesday, Microsoft announced that the Exchange Online bulk emailing rate limit is being canceled indefinitely, following negative customer feedback.

Read more of this story at Slashdot.