RSS 생중계

Phoronix's Michael Larabel reports: OpenBSD 7.6 is out this evening as another major step forward for this BSD operating system with enhanced hardware support, security improvements, updating various user-space software, and enabling other kernel enhancements. There are a ton of changes to find with the just-released OpenBSD 7.6. Some of the new OpenBSD 7.6 features include: - OpenBSD 7.6 provides initial support for Qualcomm Snapdragon X1 Elite (X1E80100) SoCs. The 7.6 release also has initial Samsung Galaxy Book4 Edge boot support in ACPI mode with OpenBSD 7.6. - ARM64 has additional CPU security mitigations with Spectre-V4 now in place on ARM64 and adding Spectre-BHB for Cortex-A57 cores. - OpenBSD 7.6 on RISC-V now supports the Milk-V Pioneer board. - OpenBSD 7.6 on AMD64 has finally implemented support for AVX-512. - Various SMP kernel improvements. You can view the full list of features and download the OpenBSD 7.6 release via OpenBSD.org.

Read more of this story at Slashdot.

BleepingComputer's Lawrence Abrams: Internet Archive's "The Wayback Machine" has suffered a data breach after a threat actor compromised the website and stole a user authentication database containing 31 million unique records. News of the breach began circulating Wednesday afternoon after visitors to archive.org began seeing a JavaScript alert created by the hacker, stating that the Internet Archive was breached. "Have you ever felt like the Internet Archive runs on sticks and is constantly on the verge of suffering a catastrophic security breach? It just happened. See 31 million of you on HIBP!," reads a JavaScript alert shown on the compromised archive.org site. The text "HIBP" refers to is the Have I Been Pwned data breach notification service created by Troy Hunt, with whom threat actors commonly share stolen data to be added to the service. Hunt told BleepingComputer that the threat actor shared the Internet Archive's authentication database nine days ago and it is a 6.4GB SQL file named "ia_users.sql." The database contains authentication information for registered members, including their email addresses, screen names, password change timestamps, Bcrypt-hashed passwords, and other internal data. Hunt says there are 31 million unique email addresses in the database, with many subscribed to the HIBP data breach notification service. The data will soon be added to HIBP, allowing users to enter their email and confirm if their data was exposed in this breach.

Read more of this story at Slashdot.

An anonymous reader quotes a report from TechCrunch: Fisker's Chapter 11 bankruptcy has hit a major snag, as the company buying the startup's remaining fleet of electric SUVs says it might not complete the purchase because of a surprising technical issue. The buyer, a New York-area leasing company called American Lease, says in a new filing that Fisker now believes there is no way to transfer the information connected to each SUV to a new server not owned by the bankrupt EV startup. Since American Lease needs that information to operate the vehicles after Fisker is dissolved, the leasing company has filed an emergency objection to the startup's liquidation plan. Fisker was expected to have that plan confirmed in bankruptcy court as early as this Wednesday. American Lease has already handed over "tens of millions of dollars" after the purchase agreement of the 3,000-plus Ocean SUVs was approved in July. These funds have been crucial because Fisker was using them to pay for the bankruptcy process. Fisker needed that money to keep itself alive long enough to settle its debts and also prepare to liquidate what it says is around $1 billion in assets that were, until recently, under control of an Austrian subsidiary that was going through its own insolvency process. [...] American Lease says in its filing that Fisker first brought up the possibility that it wouldn't be able to transfer the information to a new server on Friday, October 4, at 8 p.m. ET. And it says that this week, Fisker informed American Lease that it won't be possible at all. "[American Lease] cannot overstate the significance of this unwelcome news, conveyed to it only after it has paid [Fisker] tens of millions of dollars under the Purchase Agreement," the leasing company's lawyers write in the filing. "It is unclear at the present time what, if anything, Debtor representatives have known about the impossibility or impracticability of implementing Porting of the Purchased Vehicles, and when they learned or otherwise knew of that critical information." American Lease is asking to delay Wednesday's hearing and be allowed to perform "expedited and targeted discovery" of Fisker and its representatives to find out more about when Fisker learned of this problem.

Read more of this story at Slashdot.

80% of software engineers will need to upskill by 2027 to keep pace with generative AI's growing demands, according to Gartner. The consultancy predicts AI will transform the industry in three phases. Initially, AI tools will boost productivity, particularly for senior developers. Subsequently, "AI-native software engineering" will emerge, with most code generated by AI. Long-term, AI engineering will rise as enterprise adoption increases, requiring a new breed of professionals skilled in software engineering, data science, and machine learning.

Read more of this story at Slashdot.

An anonymous reader shares a report: Geoffrey Hinton accepted a Nobel Prize this week, recognizing the foundational work on artificial neural networks that earned him the nickname "godfather of AI." In a speech Tuesday, Professor Hinton praised one student -- alluding to OpenAI's former Chief Scientist, Ilya Sutskever -- for revolting against OpenAI's CEO. "I was particularly fortunate to have many very clever students -- much cleverer than me -- who actually made things work," said Hinton. "They've gone on to do great things. I'm particularly proud of the fact that one of my students fired Sam Altman."

Read more of this story at Slashdot.

An anonymous reader shares a report: Washington blames China's dominance of the solar industry on what are routinely dubbed "unfair trade practices." But that's just a comforting myth. China's edge doesn't come from a conspiratorial plot hatched by an authoritarian government. It hasn't been driven by state-owned manufacturers, subsidized loans to factories, tariffs on imported modules or theft of foreign technological expertise. Instead, it's come from private businesses convinced of a bright future, investing aggressively and luring global talent to a booming industry â" exactly the entrepreneurial mix that made the US an industrial powerhouse. The fall of America as a solar superpower is a tragedy of errors where myopic corporate leadership, timid financing, oligopolistic complacency and policy chaos allowed the US and Europe to neglect their own clean-tech industries. That left a yawning gap that was filled by Chinese start-ups, sprouting like saplings in a forest clearing. If rich democracies are playing to win the clean technology revolution, they need to learn the lessons of what went wrong, rather than just comfort themselves with fairy tales. To understand what happened, I visited two places: Hemlock, Michigan, a tiny community of 1,408 people that used to produce about one-quarter of the world's PV-grade polysilicon, and Leshan, China, which is now home to some of the world's biggest polysilicon factories. The similarities and differences between the towns tell the story of how the US won the 20th century's technological battle -- and how it risks losing its way in the decades ahead. [...] Meanwhile, the core questions are often almost impossible to answer. Is Tongwei's cheap electricity from a state-owned utility a form of government subsidy? What about Hemlock's tax credits protecting it from high power prices? Chinese businesses can often get cheap land in industrial parks, something that's often considered a subsidy. But does zoning US land for industrial usage count as a subsidy too? Most countries have tax credits for research and development and compete to lower their corporate tax rates to encourage investment. The factor that determines whether such initiatives are considered statist industrial policy (bad), or building a business-friendly environment (good), is usually whether they're being done by a foreign government, or our own.

Read more of this story at Slashdot.

Cutting calorie intake can lead to a leaner body -- and a longer life, an effect often chalked up to the weight loss and metabolic changes caused by consuming less food. Now, one of the biggest studies of dietary restrictions ever conducted in laboratory animals challenges the conventional wisdom about how dietary restriction boosts longevity. From a report: The study, involving nearly 1,000 mice fed low-calorie diets or subjected to regular bouts of fasting, found that such regimens do indeed cause weight loss and related metabolic changes. But other factors -- including immune health, genetics and physiological indicators of resiliency -- seem to better explain the link between cutting calories and increased lifespan. "The metabolic changes are important," says Gary Churchill, a mouse geneticist at the Jackson Laboratory in Bar Harbor, Maine, who co-led the study. "But they don't lead to lifespan extension." To outside investigators, the results drive home the intricate and individualized nature of the body's reaction to caloric restriction. "It's revelatory about the complexity of this intervention," says James Nelson, a biogerontologist at the University of Texas Health Science Center in San Antonio. The study was published today in Nature by Churchill and his co-authors, including scientists at Calico Life Sciences in South San Francisco, California, the anti-ageing focused biotech company that funded the study. Scientists have long known that caloric restriction, a regimen of long-term limits on food intake, lengthens lifespan in laboratory animals. Some studies have shown that intermittent fasting, which involves short bouts of food deprivation, can also increase longevity.

Read more of this story at Slashdot.

The man identified as Bitcoin creator Satoshi Nakamoto in a new HBO documentary has something to say: Wrong again, world. From a report: In the just-released HBO film on the history of the world's biggest digital currency -- Money Electric: The Bitcoin Mystery -- documentary filmmaker Cullen Hoback comes to the conclusion that the anonymous creator of Bitcoin was none other than a long-time member of the community and early Bitcoin developer Peter Todd. Todd dismissed the claim in the documentary, released yesterday, and denied it again when asked by The Register. "[Hoback's] evidence for me being Satoshi is the same kind of coincidence-based, circumstantial thinking that fuels conspiracies like QAnon," Todd told us in an email. "Which is ironic, given that [Hoback's] previous big project was a documentary on QAnon. He clearly didn't try to debunk his theories either." Hoback's previous project -- Q: Into the Storm -- aimed to unmask the person behind QAnon, perhaps giving him an interest in uncovering the identity of Satoshi Nakamoto. Todd, however, thinks Hoback was just trying to drum up interest in his new film. "I think [Hoback] only included the Satoshi claim as a marketing ploy: he was really creating a documentary about Bitcoin, and needed a hook to get media attention," Todd said. "He picked me to accuse mainly because I was an unlikely candidate, which helped drum up even more attention. I don't think he had any interest in finding the real truth."

Read more of this story at Slashdot.

Zoom is getting one step closer to letting AI avatars attend meetings for you. As part of a broader AI expansion, Zoom announced it will soon let you create an AI avatar of yourself that you can use to send brief messages to your team. From a report: To create a digital avatar, you'll need to record an initial video of yourself that Zoom's AI will use to make an avatar that looks -- and even sounds -- like you. From there, you can write the message you want your AI avatar to say and then have it do all the talking for you. This feature will only work with Zoom's Clips feature, allowing you to record brief video updates for your colleagues.

Read more of this story at Slashdot.

Bindgen is a widely used tool that automatically generates Rust bindings from C headers. The Rust-for-Linux project uses it to create some of the bindings between Rust code and the rest of the kernel. John Baublitz presented at Kangrejos about the improvements that he has made to the tool in order to make the generated bindings easier to use, including improved support for macros, bitfields, and enums.

9to5Mac: Following apparent photos of an M4 MacBook Pro box and a subsequent unboxing video, the unreleased model has now been offered for sale on a Russian classified ads site -- at a highly inflated price, naturally. Multiple units were advertised before the listings were removed, and it does now seem increasingly likely that the leaks are real. Apple Pro tweeted a screengrab of one of the listing, which offered what appears to be the base model 14-inch M4 MacBook Pro, with the previously reported specs of 16GB unified memory, 512GB SSD, and three Thunderbolt 4 ports. We've also seen Geekbench results for a machine identified as "Mac 16,1" with performances in line with the reported specs. Rather than a one-off leak, it has been claimed that there are some 200 units out there. The ad on Avito was asking 720,000 rubles, which is around $7,400.

Read more of this story at Slashdot.

The greatest tennis players in the world will be left to rage against a machine after any tight line calls at Wimbledon next year as the All England Club will break with tradition by removing line judges from all courts during the championships for the first time in its 147-year history. From a report: From 2025 onwards live electronic line calling (ELC) will be used on all courts in both the main draw at the All England Club and the qualifying tournament off-site in Roehampton. The new technology was successfully tested during this year's championships. Wimbledon's chief executive, Sally Bolton, said: "The decision to introduce Live Electronic Line Calling at the championships was made following a significant period of consideration and consultation." Bolton added: "Having reviewed the results of the testing this year, we consider the technology to be sufficiently robust and the time is right to take this important step in seeking maximum accuracy in our officiating. For the players, it will offer them the same conditions they have played under at a number of other events on tour."

Read more of this story at Slashdot.

Nintendo has a new piece of game hardware coming in 2024. It's not the Switch 2, it's the Nintendo Sound Clock: Alarmo, a new motion-sensing alarm clock that will rouse you from sleep with the sounds of video games like Super Mario Odyssey, The Legend of Zelda: Breath of the Wild, and more. Polygon adds: The $99.99 alarm clock includes 35 wake-up "scenes" from Nintendo Switch games. Users will "experience immersive sounds and music" from those scenes, which are styled to match Nintendo's franchises. Alarmo will also track users' sleep and motion; Records will keep track of how much you move around, similar to the Pokemon Go Plus Plus sleep tracker that Nintendo released for Pokemon Go/Pokemon Sleep (and the long-promised, but seemingly abandoned QOL project). Nintendo Sound Clock: Alarmo will feature various alarm settings, Nintendo said in a news release. The clock's Steady Mode features an alarm that will gradually get more intense the longer you stay in bed. Gentle Mode offers "a more consistent intensity level." And Button Mode is a classic snooze button mode, where you've gotta smack Alarmo to shut it up. Nintendo's new interactive, motion-sensing alarm clock also features "sleepy sounds" -- soothing music that will hopefully lull you to sleep at night.

Read more of this story at Slashdot.

OpenAI said a group with apparent ties to China tried to carry out a phishing attack on its employees, reigniting concerns that bad actors in Beijing want to steal sensitive information from top US artificial intelligence companies. From a report: The AI startup said Wednesday that a suspected China-based group called SweetSpecter posed as a user of OpenAI's chatbot ChatGPT earlier this year and sent customer support emails to staff. The emails included malware attachments that, if opened, would have allowed SweetSpecter to take screenshots and exfiltrate data, OpenAI said, but the attempt was unsuccessful. "OpenAI's security team contacted employees who were believed to have been targeted in this spear phishing campaign and found that existing security controls prevented the emails from ever reaching their corporate emails," OpenAI said. The disclosure highlights the potential cybersecurity risks for leading AI companies as the US and China are locked in a high-stakes battle for artificial intelligence supremacy. In March, for example, a former Google engineer was charged with stealing AI trade secrets for a Chinese firm.

Read more of this story at Slashdot.

AmiMoJo writes: A treaty finalized by the UK may bring about the end of the .io domain. Last week, the British government announced that it has agreed to give up ownership of the Chagos Islands, a territory in the Indian Ocean it has controlled since 1814 -- relinquishing the .io domain with it. The Internet Assigned Numbers Authority (IANA) has a process for retiring old country code domains within five years (with the possibility for extensions). The IANA established this rule after the Soviet Union's .su domain lingered after its collapse, becoming a domain commonly used among cybercriminals. Since then, IANA has also had to retire the .yu domain previously used for Yugoslavia, but it remained operational for years following the country's breakup while government websites transitioned to new domains. And while the independent Solomon Islands does have the domain name .sb, where 'B' stands for how it used to be a British protectorate, that domain was registered decades after it achieved independence. The UK still has the inactive .gb domain as well, but it's considering getting rid of it.

Read more of this story at Slashdot.

The Julia project has released version 1.11.0. A separate blog post covers some of the highlights. The release includes a number of helpful features.

In previous Julia versions, there was no "programmatic way" of knowing if an unexported name was considered part of the public API or not. Instead, the guideline was basically that if it was not in the manual then it was not public which was a bit underwhelming. To remedy that, there is now a public keyword in Julia that can be used to indicate that an unexported name is part of the public API.

Three scientists won the 2024 Nobel Prize in Chemistry on Wednesday for their groundbreaking work in predicting and designing protein structures, the Royal Swedish Academy of Sciences announced in Stockholm. David Baker of the University of Washington shares the prize with Demis Hassabis and John Jumper of Google DeepMind. Baker pioneered the creation of novel proteins, while Hassabis and Jumper developed AlphaFold, an AI model that predicts protein structures from amino acid sequences. The laureates will split the 11 million Swedish kronor ($1 million) award for their contributions to computational protein design and structure prediction. Baker's team has produced proteins with applications in medicine and materials science since his initial breakthrough in 2003. Hassabis and Jumper's AlphaFold, announced in 2020, has predicted structures for nearly all 200 million known proteins. "We glimpsed at the beginning that it might be possible to create a whole new world of proteins that address a lot of the problems faced by humans in the 21st century," Baker said at a press briefing. "Now it's becoming possible," Heiner Linke, chair of the Nobel chemistry committee, called the discoveries "spectacular," noting they fulfilled a 50-year-old dream of predicting protein structures from amino acid sequences. The breakthroughs have wide-ranging implications, from understanding antibiotic resistance to developing enzymes that decompose plastic. Over 2 million researchers worldwide have already utilized AlphaFold in various applications.

Read more of this story at Slashdot.

Security updates have been issued by AlmaLinux (firefox, mod_jk, and thunderbird), Debian (apache2 and firefox-esr), Fedora (crosswords, logiops, p7zip, and perl-App-cpanminus), Red Hat (.NET 6.0, firefox, git, kernel, kernel-rt, openssl, and thunderbird), SUSE (buildah, json-lib, kernel, Mesa, mozjs78, pgadmin4, podman, podofo, qatlib, redis7, roundcubemail, rusty_v8, and seamonkey), and Ubuntu (dotnet6, dotnet8, nginx, and ruby-webrick).

In a new 32-page filing (PDF), the Department of Justice indicated that it was considering a possible breakup of Google as an antitrust remedy for its search and advertising monopoly. The remedies necessary to "prevent and restrain monopoly maintenance could include contract requirements and prohibitions; non-discrimination product requirements; data and interoperability requirements; and structural requirements," the department said in the filing. CNBC reports: The DOJ also said it was "considering behavioral and structural remedies that would prevent Google from using products such as Chrome, Play, and Android to advantage Google search and Google search-related products and features -- including emerging search access points and features, such as artificial intelligence -- over rivals or new entrants." Additionally, the DOJ suggested limiting or prohibiting default agreements and "other revenue-sharing arrangements related to search and search-related products." That would include Google's search position agreements with Apple's iPhone and Samsung devices -- deals that cost the company billions of dollars a year in payouts. The agency suggested one way to do this is requiring a "choice screen," which could allow users to pick from other search engines. Such remedies would end "Google's control of distribution today" and ensure "Google cannot control the distribution of tomorrow."

Read more of this story at Slashdot.

Researchers at BitEnergy AI, Inc. have developed Linear-Complexity Multiplication (L-Mul), a technique that reduces AI model power consumption by up to 95% by replacing energy-intensive floating-point multiplications with simpler integer additions. This method promises significant energy savings without compromising accuracy, but it requires specialized hardware to fully realize its benefits. Decrypt reports: L-Mul tackles the AI energy problem head-on by reimagining how AI models handle calculations. Instead of complex floating-point multiplications, L-Mul approximates these operations using integer additions. So, for example, instead of multiplying 123.45 by 67.89, L-Mul breaks it down into smaller, easier steps using addition. This makes the calculations faster and uses less energy, while still maintaining accuracy. The results seem promising. "Applying the L-Mul operation in tensor processing hardware can potentially reduce 95% energy cost by element wise floating point tensor multiplications and 80% energy cost of dot products," the researchers claim. Without getting overly complicated, what that means is simply this: If a model used this technique, it would require 95% less energy to think, and 80% less energy to come up with new ideas, according to this research. The algorithm's impact extends beyond energy savings. L-Mul outperforms current 8-bit standards in some cases, achieving higher precision while using significantly less bit-level computation. Tests across natural language processing, vision tasks, and symbolic reasoning showed an average performance drop of just 0.07% -- a negligible tradeoff for the potential energy savings. Transformer-based models, the backbone of large language models like GPT, could benefit greatly from L-Mul. The algorithm seamlessly integrates into the attention mechanism, a computationally intensive part of these models. Tests on popular models such as Llama, Mistral, and Gemma even revealed some accuracy gain on certain vision tasks. At an operational level, L-Mul's advantages become even clearer. The research shows that multiplying two float8 numbers (the way AI models would operate today) requires 325 operations, while L-Mul uses only 157 -- less than half. "To summarize the error and complexity analysis, L-Mul is both more efficient and more accurate than fp8 multiplication," the study concludes. But nothing is perfect and this technique has a major achilles heel: It requires a special type of hardware, so the current hardware isn't optimized to take full advantage of it. Plans for specialized hardware that natively supports L-Mul calculations may be already in motion. "To unlock the full potential of our proposed method, we will implement the L-Mul and L-Matmul kernel algorithms on hardware level and develop programming APIs for high-level model design," the researchers say.

Read more of this story at Slashdot.