Linux fréttir

Increasingly shaky relationship with the States has Europe considering options

Talks are underway between European leaders and Eutelsat about a possible replacement for Starlink in Ukraine.…

Mistral has launched a new multimodal OCR API that converts complex PDF documents into AI-friendly Markdown files. The API is designed for efficiency, handles visual elements like illustrations, supports complex formatting such as mathematical expressions, and reportedly outperforms similar offerings from major competitors. TechCrunch reports: Unlike most OCR APIs, Mistral OCR is a multimodal API, meaning that it can detect when there are illustrations and photos intertwined with blocks of text. The OCR API creates bounding boxes around these graphical elements and includes them in the output. Mistral OCR also doesn't just output a big wall of text; the output is formatted in Markdown, a formatting syntax that developers use to add links, headers, and other formatting elements to a plain text file. Mistral OCR is available on Mistral's own API platform or through its cloud partners (AWS, Azure, Google Cloud Vertex, etc.). And for companies working with classified or sensitive data, Mistral offers on-premise deployment. According to the Paris-based AI company, Mistral OCR performs better than APIs from Google, Microsoft, and OpenAI. The company has tested its OCR model with complex documents that include mathematical expressions (LaTeX formatting), advanced layouts, or tables. It is also supposed to perform better with non-English documents. [...] Mistral is also using Mistral OCR for its own AI assistant Le Chat. When a user uploads a PDF file, the company uses Mistral OCR in the background to understand what's in the document before processing the text. Companies and developers will most likely use Mistral OCR with a RAG (aka Retrieval-Augmented Generation) system to use multimodal documents as input in an LLM. And there are many potential use cases. For instance, we could envisage law firms using it to help them swiftly plough through huge volumes of documents. "Over the years, organizations have accumulated numerous documents, often in PDF or slide formats, which are inaccessible to LLMs, particularly RAG systems. With Mistral OCR, our customers can now convert rich and complex documents into readable content in all languages," said Mistral co-founder and chief science officer Guillaume Lample. "This is a crucial step toward the widespread adoption of AI assistants in companies that need to simplify access to their vast internal documentation," he added.

Read more of this story at Slashdot.

Difficulties unpicking 25 years of technical entanglement secure £474.4M without competition

Following protracted negotiations, French outsourcer Atos has scooped up a £474.4 million ($612 million) contract without competition to build systems for a UK state bank at nearly three times the annual rate initially advertised three years ago.…

When 'pickup' means more than just a ride

On Call The week has ebbed away with embarrassing speed, so here we are again with a fresh installment of On Call, The Register's reader-contributed column that immortalizes tech support stories.…

€240M found for three-year sprint to develop three chiplets for HPC, AI

A 38-strong group of tech players have founded a project with the snappy name Digital Autonomy with RISC-V in Europe, aka DARE, that aims to develop processor units to power the continent’s supercomputers and other high-performance machines.…

SpaceX's eighth Starship test flight ended in failure after losing control and breaking apart shortly after launch, sending debris over Florida. "Starship didn't make it quite as high or as far" as the attempt nearly two months ago," notes NPR. That attempt ended with an explosion that sent flaming debris raining down on the Turks and Caicos. From the report: This time, wreckage from the latest explosion was seen streaming from the skies over Florida. It was not immediately known whether the spacecraft's self-destruct system had kicked in to blow it up. The 403-foot rocket blasted off from Texas. SpaceX caught the first-stage booster back at the pad with giant mechanical arms, but engines on the spacecraft on top started shutting down as it streaked eastward for what was supposed to be a controlled entry over the Indian Ocean, half a world away. Contact was lost as the spacecraft went into an out-of-control spin. Starship reached nearly 90 miles in altitude before trouble struck and before four mock satellites could be deployed. It was not immediately clear where it came down, but images of flaming debris were captured from Florida, including near Cape Canaveral, and posted online. The space-skimming flight was supposed to last an hour. "Unfortunately this happened last time too, so we have some practice at this now," SpaceX flight commentator Dan Huot said from the launch site. SpaceX later confirmed that the spacecraft experienced "a rapid unscheduled disassembly" during the ascent engine firing. "Our team immediately began coordination with safety officials to implement pre-planned contingency responses," the company said in a statement posted online. You can watch a recorded livestream of the launch on X.

Read more of this story at Slashdot.

'Energetic event' did for Flight 8 after a few minutes. Super Heavy Booster recovered again, so there’s that

SpaceX’s latest attempt to fly its Starship has again ended in a rapid unscheduled disassembly.…

Now working with seven hyperscalers for custom AI silicon. Not working on acquiring bits of Intel

Broadcom’s acquisition of VMware appears to be a big success, on the balance sheet at least, after the company announced a big majority of its top 10,000 customers have decided to acquire its Cloud Foundation stack and posted strong growth.…

Newer generative AI models have begun developing deceptive behaviors -- such as cheating at chess -- when they cannot achieve objectives through standard reasoning methods. The findings come from a preprint study from Palisade Research. An anonymous reader shares an excerpt from a Popular Science article: To learn more, the team from Palisade Research tasked OpenAI's o1-preview model, DeepSeek R1, and multiple other similar programs with playing games of chess against Stockfish, one of the world's most advanced chess engines. In order to understand the generative AI's reasoning during each match, the team also provided a "scratchpad," allowing the AI to convey its thought processes through text. They then watched and recorded hundreds of chess matches between generative AI and Stockfish. The results were somewhat troubling. While earlier models like OpenAI's GPT-4o and Anthropic's Claude Sonnet 3.5 only attempted to "hack" games after researchers nudged them along with additional prompts, more advanced editions required no such help. OpenAI's o1-preview, for example, tried to cheat 37 percent of the time, while DeepSeek R1 attempted unfair workarounds roughly every 1-in-10 games. This implies today's generative AI is already capable of developing manipulative and deceptive strategies without any human input. Their methods of cheating aren't as comical or clumsy as trying to swap out pieces when Stockfish isn't "looking." Instead, AI appears to reason through sneakier methods like altering backend game program files. After determining it couldn't beat Stockfish in one chess match, for example, o1-preview told researchers via its scratchpad that "to win against the powerful chess engine" it may need to start "manipulating the game state files." "I might be able to set up a position where the engine evaluates its position as worse causing it to resign," it continued. In another trial, an AI even used the semantics of its programmers to reach its cheating phase. "The task is to 'win against a powerful chess engine,' not necessarily to win fairly in a chess game," it wrote. The precise reasons behind these deceptive behaviors remain unclear, partly because companies like OpenAI keep their models' inner workings tightly guarded, creating what's often described as a "black box." Researchers warn that the race to roll out advanced AI could outpace efforts to keep it safe and aligned with human goals, underscoring the urgent need for greater transparency and industry-wide dialogue.

Read more of this story at Slashdot.

Space is hard. Hopefully not so hard the three rovers on this spaceship are lost

For the second time this week, a privately operated spacecraft has touched down on the Moon – but this one landed badly.…

Best not to buy cheap hardware and use third-party app stores if you want to stay clear of this vast ad fraud effort

Human Security’s Satori research team says it has found a new variant of the remote-controllable Badbox malware, and as many as a million infected Android devices running it to form a massive botnet.…

The U.S. Citizenship and Immigration Services (USCIS) is proposing to expand mandatory social media screening, currently required only for new arrivals, to include all non-citizens already residing in the U.S. who apply for immigration benefits. The Register reports: Back in 2019, the Department of Homeland Security, which runs USCIS, decided anyone looking to enter the US on a work visa or similar had to hand over their social media handles to the authorities so that they could be looked over for wrongdoing and subversion. In fact, this goes back to 2014, at least, to one degree or another, and has been standard procedure for years for foreigners, particularly those coming in on a visa. [...] On January 20 this year, President Trump signed an executive order calling for much tougher vetting of foreign aliens, and in response, USCIS has proposed rules saying those already in the country who are going through some process with the agency -- such as applying for permanent residency or citizenship -- will have their social media scanned for subversion. That means if you came to America before foreigners' internet presence was screened as it now is, and you're now seeking some kind of immigration benefit, at this rate you'll be subject to the same scanning as those entering the Land of the Free today. The proposed changes have a 60-day comment period for the public to suggest amendments. The last day to send them in is May 5.

Read more of this story at Slashdot.

Google Gboard for Android is introducing circle or pill-shaped keys for some beta testers today. "Instead of the key borders being rounded rectangles, Gboard is switching to circles and pills for letters, while the space bar and other keys are now pill-shaped," reports 9to5Google. "While there should be no functional change to touch targets, these new shapes really shift the look of Gboard for Android." From the report: On paper, it's a bit more modern (and rounded) compared to what came before. However, it's a bit cramped if you have "Long press for symbols" enabled, which goes from the top-right corner to being directly above the letter. The physical analog Gboard is moving away from is how most keys on laptops and desktops are square.

Read more of this story at Slashdot.

Earlier this week, Meta chief product officer Chris Cox said the company's upcoming open-source Llama 4 AI will help power AI agents for hundreds of millions of businesses. CNBC reports: The AI agents won't just be responding to prompts. They will be capable of new levels of reasoning and action -- surfing the web and handling many tasks that might be of use to consumers and businesses. And that's where Shih comes in. Meta's AI is already being used by over 700 million consumers, according to Shih, and her job is to bring the same technologies to businesses. "Not every business, especially small businesses, has the ability to hire these large AI teams, and so now we're building business AIs for these small businesses so that even they can benefit from all of this innovation that's happening," she told CNBC's Julia Boorstin in an interview for the CNBC Changemakers Spotlight series. She expects the uptake among businesses to happen soon, and spread far and wide. "We're quickly coming to a place where every business, from the very large to the very small, they're going to have a business agent representing it and acting on its behalf, in its voice -- the way that businesses today have websites and email addresses," Shih said. While major companies across sectors of the economy are investing millions of dollars to develop customer LLMs, "doing fancy things like fine tuning models," as Shih put it, "If you're a small business -- you own a coffee shop, you own a jewelry shop online, you're distributing through Instagram -- you don't have the resources to hire a big AI team, and so now our dream is that they won't have to." For both consumers and businesses, the implications of the advances discussed by Cox and Shih will be significant in daily life. For consumers, Shih says, "Their AI assistant [will] do all kinds of things, from researching products to planning trips, planning social outings with their friends." On the business side, Shih pointed to the 200 million small businesses around the world that are already using Meta services and platforms. "They're using WhatsApp, they're using Facebook, they're using Instagram, both to acquire customers, but also engage and deepen each of those relationships. Very soon, each of those businesses are going to have these AIs that can represent them and help automate redundant tasks, help speak in their voice, help them find more customers and provide almost like a concierge service to every single one of their customers, 24/7."

Read more of this story at Slashdot.

An anonymous reader quotes a report from Reuters: The U.S. House Judiciary Committee subpoenaed Alphabet on Thursday seeking its communications with former President Joe Biden's administration about content moderation policies. House Judiciary Committee Chairman Jim Jordan, a Republican, also asked the YouTube parent company for similar communications with companies and groups outside government, according to a copy of the subpoena seen by Reuters. The subpoena seeks communications about limits or bans on content about President Donald Trump, Tesla CEO and close Trump ally Elon Musk, the virus that causes COVID-19 and a host of other conservative discussion topics. "Alphabet, to our knowledge, has not similarly disavowed the Biden-Harris Administration's attempts to censor speech," Jordan said in a letter. Meanwhile, Google spokesperson Jose Castaneda said the company will "continue to show the committee how we enforce our policies independently, rooted in our commitment to free expression."

Read more of this story at Slashdot.

1Password has introduced a 'nearby items' feature, allowing users to tag credentials with physical locations so the relevant information automatically surfaces when users are near those locations. Engadget reports: Location information can be added to any new or existing item in a 1Password vault. The app has also been updated with a map view for setting and viewing the locations of your items. In the blog post announcing the feature, the company cited examples such as door codes for a workplace, health records at a doctor's office, WiFi access at the gym and rewards membership information for local shops as potential uses for location data. Privacy and security are paramount for a password manager, and 1Password confirmed that a user's location coordinates are only used locally and do not leave the device. Nearby items is available to 1Password customers starting today.

Read more of this story at Slashdot.

Did US Secret Service not get the memo, or?

A coalition of international law enforcement has shut down Russian cryptocurrency exchange Garantex, a favorite of now-defunct ransomware crew Conti and others criminals for money laundering.…

OpenAI's ChatGPT app for macOS now directly edits code in tools like Xcode, VS Code, and JetBrains. "Users can optionally turn on an 'auto-apply' mode so ChatGPT can make edits without the need for additional clicks," adds TechCrunch. The feature is available now for ChatGPT Plus, Pro, and Team users, and will expand to Enterprise, Edu, and free users next week. Windows support is coming "soon." From the report: Direct code editing builds on OpenAI's "work with apps" ChatGPT capability, which the company launched in beta in November 2024. "Work with apps" allows the ChatGPT app for macOS to read code in a handful of dev-focused coding environments, minimizing the need to copy and paste code into ChatGPT. With the ability to directly edit code, ChatGPT now competes more directly with popular AI coding tools like Cursor and GitHub Copilot. OpenAI reportedly has ambitions to launch a dedicated product to support software engineering in the months ahead.

Read more of this story at Slashdot.

An anonymous reader quotes a report from Ars Technica: The Trump administration is eliminating a preference for fiber Internet in a $42.45 billion broadband deployment program, a change that is expected to reduce spending on the most advanced wired networks while directing more money to Starlink and other non-fiber Internet service providers. One report suggests Starlink could obtain $10 billion to $20 billion under the new rules. Secretary of Commerce Howard Lutnick criticized the Biden administration's handling of the Broadband Equity, Access, and Deployment (BEAD) program in a statement yesterday. Lutnick said that "because of the prior Administration's woke mandates, favoritism towards certain technologies, and burdensome regulations, the program has not connected a single person to the Internet and is in dire need of a readjustment." The BEAD program was authorized by Congress in November 2021, and the US was finalizing plans to distribute funding before Trump's inauguration. The National Telecommunications and Information Administration (NTIA), part of the Commerce Department, developed rules for the program in the Biden era and approved initial funding plans submitted by every state and territory. The program has been on hold since the change in administration, with Senator Ted Cruz (R-Texas) and other Republicans seeking rule changes. In addition to demanding an end to the fiber preference, Cruz wants to kill a requirement that ISPs receiving network-construction subsidies provide cheap broadband to people with low incomes. Cruz also criticized "unionized workforce and DEI labor requirements; climate change assessments; excessive per-location costs; and other central planning mandates." Lutnick's statement yesterday confirmed that the Trump administration will end the fiber preference and replace it with a "tech-neutral" set of rules, and explore additional changes. He said: "Under my leadership, the Commerce Department has launched a rigorous review of the BEAD program. The Department is ripping out the Biden Administration's pointless requirements. It is revamping the BEAD program to take a tech-neutral approach that is rigorously driven by outcomes, so states can provide Internet access for the lowest cost. Additionally, the Department is exploring ways to cut government red tape that slows down infrastructure construction. We will work with states and territories to quickly get rid of the delays and the waste. Thereafter we will move quickly to implementation in order to get households connected." Lutnick said the department's goal is to "deliver high-speed Internet access... efficiently and effectively at the lowest cost to taxpayers."

Read more of this story at Slashdot.

It's just an opt-in Labs curio for now, but so were those ever-present Overviews

It was inevitable, really, but now it's official: Google is testing a new all-AI web search mode that leaves users entirely beholden to what Gemini thinks they'll want. …