Google

Google Says It Might Have Deleted Your Maps Timeline Data (arstechnica.com) 14

Google has confirmed that a technical issue has permanently deleted location history data for numerous users of its Maps application, with no recovery possible for most affected customers. The problem emerged after Google transitioned its Timeline feature from cloud to on-device storage in 2024 to enhance privacy protections. Users began reporting missing historical location data on support forums and social media platforms in recent weeks. "This is the result of a technical issue and not user error or an intentional change," said a Google spokesperson. Only users who manually enabled encrypted cloud backups before the incident can recover their data, according to Google. The company began shifting location storage policies in 2023, initially stopping collection of sensitive location data including visits to abortion clinics and domestic violence shelters.
IBM

IBM Cuts Thousands of Jobs, Cloud Classic Unit Hit Hard: Report (theregister.com) 49

IBM is laying off thousands of employees across the United States, with approximately 25% of staff at its Cloud Classic operation affected, The Register reports, citing a source. "Concrete numbers are being kept private," a source told the publication. "It is in the thousands."

Staff reductions have occurred in Raleigh, North Carolina; New York; Dallas, Texas; and California, the report said. Affected departments include consulting, corporate social responsibility, cloud infrastructure, sales, and internal systems teams. The report adds: With regard to IBM Cloud Classic -- the infrastructure-as-a-service (IaaS) outfit offering built on IBM's 2013 acquisition of SoftLayer -- another source told us: "It's a resource action. I don't know how many people are in IaaS classic. They don't typically make that information easy to find. What I can say is that they have been making a lot of changes to shift employment to India as much as possible."

A third source, newly let go by Big Blue, said it was fair to characterize this a layoff. "Everyone I know that was affected, myself included, was simply offered a separation agreement," this individual said, estimating that 10 percent of the Cloud group (which is not the same as Cloud Classic) has been let go.

IT

PCI Express 7.0's Blazing Speeds Are Nearly Here, But PCIe 6 is Still Vapor (pcworld.com) 63

An anonymous reader shares a report: PCI Express 7 is nearing completion, the PCI Special Interest Group said, and the final specification should be released later this year. PCI Express 7, the backbone of the modern motherboard, is at the stage 0.9, which the PCI-SIG characterizes as the "final draft" of the specification. The technology was at version 0.5 a year ago, almost to the day, and originally authored in 2022.

The situation remains the same, however. While modern PC motherboards are stuck on PCI Express 5.0, the specification itself moves ahead. PCI Express has doubled the data rate about every three years, from 64 gigtransfers per second in PCI Express 6.0 to the upcoming 128 gigatransfers per second in PCIe 7. (Again, it's worth noting that PCIe 6.0 exists solely on paper.) Put another way, PCIe 7 will deliver 512GB/s in both directions, across a x16 connection.

It's worth noting that the PCI-SIG doesn't see PCI Express 7 living inside the PC market, at least not initially. Instead, PCIe 7 is expected to be targeted at cloud computing, 800-gigabit Ethernet and, of course, artificial intelligence. It will be backwards-compatible with the previous iterations of PCI Express, the SIG said.

AI

AI Crawlers Haven't Learned To Play Nice With Websites (theregister.com) 57

SourceHut, an open-source-friendly git-hosting service, says web crawlers for AI companies are slowing down services through their excessive demands for data. From a report: "SourceHut continues to face disruptions due to aggressive LLM crawlers," the biz reported Monday on its status page. "We are continuously working to deploy mitigations. We have deployed a number of mitigations which are keeping the problem contained for now. However, some of our mitigations may impact end-users."

SourceHut said it had deployed Nepenthes, a tar pit to catch web crawlers that scrape data primarily for training large language models, and noted that doing so might degrade access to some web pages for users. "We have unilaterally blocked several cloud providers, including GCP [Google Cloud] and [Microsoft] Azure, for the high volumes of bot traffic originating from their networks," the biz said, advising administrators of services that integrate with SourceHut to get in touch to arrange an exception to the blocking.

EU

Dutch Parliament Calls For End To Dependence On US Software Companies (yahoo.com) 106

The Dutch parliament approved motions urging the government to reduce reliance on U.S. software companies by developing a sovereign cloud platform and reconsidering contracts with American firms. Reuters reports: While such initiatives have foundered in the past due to a lack of viable European alternatives, lawmakers said changing relations with the United States under the presidency of Donald Trump have given the issue fresh urgency. "The question we as Europeans must ask ourselves is: do we feel comfortable with people like Trump, (Meta CEO Mark) Zuckerberg and (X owner Elon) Musk ruling over our data?" said Marieke Koekkoek of the pro-European Volt party, who authored one of the eight motions, in an email to Reuters.

In addition to launching a sovereign cloud services platform, the motions called on the government to re-examine a decision to use Amazon's web services for the Netherlands' internet domain hosting, and to develop alternatives to U.S. software and preferential treatment for European firms in public tenders. [...] Bert Hubert, a Dutch technology expert who has advocated for reducing dependency on the U.S., said: "This is only the first step in potentially doing something." But he said one important outcome would be forcing agencies to publicly report on risks related to their reliance on U.S. cloud firms. "With the advent of Trump 2.0, it has become clear that this is not something you can harmlessly sign off on," he said.

Businesses

Alphabet Back In Talks To Buy Wiz For $30 Billion (yahoo.com) 14

Google's parent company Alphabet is reportedly in talks to acquire cybersecurity startup Wiz for approximately $30 billion. Last July, negotiations had advanced on a $23 billion deal, but the talks were put on hold to prioritize Wiz's IPO. Around the same time, Alphabet also walked away from a potential acquisition of online marketing software company HubSpot. Reuters reports: The startup provides cloud-based cybersecurity solutions powered by artificial intelligence that help companies identify and remove critical risks on cloud platforms. A buyout of this size will most likely face regulatory scrutiny as tech giants are kept under close watch for possible monopolistic practices.

If the deal goes through, it could help Alphabet tap into the cybersecurity industry and expand its booming cloud infrastructure segment, which generated more than $43 billion in revenue last year. Wiz was last valued at $12 billion in a private funding round in May 2024.

Businesses

Remote Working Saved Zillow Money, Helped Recruiting, and Maintained Productivity (seattletimes.com) 40

Zillow CEO Jeremy Wacksman "recently told Entrepreneur magazine that almost five years of remote work has 'been fantastic for us,'" writes the Seattle Times. Zillow shifted to allowing people to work fully remote during the pandemic. It's been a recruiting and retention tool for Zillow as they "now see four times the number of job applicants for every job we have versus what we did before the pandemic," Wacksman said.

While Zillow still lists its corporate headquarters as Seattle, the company bills itself as "cloud-headquartered," with remote workers and satellite offices. Wacksman's comments are backed by serious real estate moves the company has made over the past five years. An annual report detailing Zillow's financial results for 2024 shows its Seattle headquarters and offices across the country are shrinking. In 2019, Zillow had 386,275 square feet of office space in Seattle after steadily gobbling up floors of the Russell Investments Center downtown over the prior five years. The company reported it had 113,470 square feet in Seattle at the end of 2024... The company has drastically cut costs by shedding offices. Zillow's total leasing costs reached $54 million in 2022 and dropped to $34 million last year... It expects those costs to decrease even further, to $18 million by 2029. Zillow is also taking advantage of subleasing some of its office space and expects $26 million in sublease income between 2025 and 2030...

Zillow's financial results from last year suggest the workforce has been productive while logging in from home. The company reported Tuesday that it beat Wall Street expectations for the last three months of 2024 with a quarterly revenue of $554 million. Wacksman said in a news release Tuesday that 2024 was a "remarkable year for Zillow," as it reached its goal of double-digit revenue growth.

Privacy

Everything You Say To Your Echo Will Be Sent To Amazon Starting On March 28 (arstechnica.com) 43

An anonymous reader quotes a report from Ars Technica: In an email sent to customers today, Amazon said that Echo users will no longer be able to set their devices to process Alexa requests locally and, therefore, avoid sending voice recordings to Amazon's cloud. Amazon apparently sent the email to users with "Do Not Send Voice Recordings" enabled on their Echo. Starting on March 28, recordings of everything spoken to the Alexa living in Echo speakers and smart displays will automatically be sent to Amazon and processed in the cloud.

Attempting to rationalize the change, Amazon's email said: "As we continue to expand Alexa's capabilities with generative AI features that rely on the processing power of Amazon's secure cloud, we have decided to no longer support this feature." One of the most marketed features of Alexa+ is its more advanced ability to recognize who is speaking to it, a feature known as Alexa Voice ID. To accommodate this feature, Amazon is eliminating a privacy-focused capability for all Echo users, even those who aren't interested in the subscription-based version of Alexa or want to use Alexa+ but not its ability to recognize different voices.

[...] Amazon said in its email today that by default, it will delete recordings of users' Alexa requests after processing. However, anyone with their Echo device set to "Don't save recordings" will see their already-purchased devices' Voice ID feature bricked. Voice ID enables Alexa to do things like share user-specified calendar events, reminders, music, and more. Previously, Amazon has said that "if you choose not to save any voice recordings, Voice ID may not work." As of March 28, broken Voice ID is a guarantee for people who don't let Amazon store their voice recordings.
Amazon's email continues: "Alexa voice requests are always encrypted in transit to Amazon's secure cloud, which was designed with layers of security protections to keep customer information safe. Customers can continue to choose from a robust set of controls by visiting the Alexa Privacy dashboard online or navigating to More - Alexa Privacy in the Alexa app."

Further reading: Google's Gemini AI Can Now See Your Search History
The Almighty Buck

Tariffs Are Proving 'Big Headache' For Tech Giants, Says Foxconn (ft.com) 213

The US government's tariff announcements have become a "big headache" for technology companies such as iPhone maker Apple and cloud service provider Amazon, their manufacturing partner Foxconn said on Friday, in a rare public admission of the disruption caused by President Donald Trump's erratic trade policy. Financial Times: "The issue of tariffs is something that is giving the CEOs of our customers a big headache now," chief executive Young Liu told investors on an earnings call. "Judging by the attitude and the approach we see the US government taking towards tariffs, it is very, very hard to predict how things will develop over the next year. So we can only concentrate on doing well what we can control."

Liu said the company's customers were "one after another" hatching plans for co-operating with Foxconn on manufacturing in the US. He declined to give details as those plans were not yet finalised, but said there should be "more and more" manufacturing in the US.

Google

UK Investigation Says Apple, Google Hampering Mobile Browser Competition 14

Britain's competition watchdog has concluded that Apple and Google are stifling competition in the UK mobile browser market, following an investigation by the Competition and Markets Authority (CMA). The inquiry found Apple's iOS policies particularly restrictive, requiring all browsers to use its WebKit engine while giving Safari preferential access to features.

Apple's practice of pre-installing Safari as the default browser also reduces awareness of alternatives, despite allowing users to change defaults. Google faces similar criticism for pre-installing Chrome on most Android devices, though investigators noted both companies have recently taken steps to facilitate browser switching. The probe identified Apple's revenue-sharing arrangement with Google -- which pays a significant share of search revenue to be the default iPhone search engine -- as "significantly reducing their financial incentives to compete."
Earth

Geothermal Could Power Nearly All New Data Centers Through 2030 (techcrunch.com) 26

An anonymous reader quotes a report from TechCrunch: There's a power crunch looming as AI and cloud providers ramp up data center construction. But a new report suggests that a solution lies beneath their foundations. Advanced geothermal power could supply nearly two-thirds of new data center demand by 2030, according to an analysis by the Rhodium Group. The additions would quadruple the amount of geothermal power capacity in the U.S. -- from 4 gigawatts to about 16 gigawatts -- while costing the same or less than what data center operators pay today. In the western U.S., where geothermal resources are more plentiful, the technology could provide 100% of new data center demand. Phoenix, for example, could add 3.8 gigawatts of data center capacity without building a single new conventional power plant.

Geothermal resources have enormous potential to provide consistent power. Historically, geothermal power plants have been limited to places where Earth's heat seeps close to the surface. But advanced geothermal techniques could unlock 90 gigawatts of clean power in the U.S. alone, according to the U.S. Department of Energy. [...] Because geothermal power has very low running costs, its price is competitive with data centers' energy costs today, the Rhodium report said. When data centers are sited similarly to how they are today, a process that typically takes into account proximity to fiber optics and major metro areas, geothermal power costs just over $75 per megawatt hour. But when developers account for geothermal potential in their siting, the costs drop significantly, down to around $50 per megawatt hour.

The report assumes that new generating capacity would be "behind the meter," which is what experts call power plants that are hooked up directly to a customer, bypassing the grid. Wait times for new power plants to connect to the grid can stretch on for years. As a result, behind the meter arrangements have become more appealing for data center operators who are scrambling to build new capacity.

Microsoft

Microsoft is Replacing Remote Desktop With Its New Windows App (theverge.com) 52

Microsoft is ending support of its Remote Desktop app for Windows on May 27th. From a report: If you use the Remote Desktop app to connect to Windows 365, Azure Virtual Desktop, or Microsoft Dev Box machines then you'll have to transition to the Windows app instead.

The new Windows app, which launched in September, includes multimonitor support, dynamic display resolutions, and easy access to cloud PCs and virtual desktops. Microsoft says "connections to Windows 365, Azure Virtual Desktop, and Microsoft Dev Box via the Remote Desktop app from the Microsoft Store will be blocked after May 27th, 2025."

Microsoft

Microsoft To Launch Xbox Handheld in 2025, Next-Gen Consoles in 2027 (windowscentral.com) 33

Microsoft plans to launch an Xbox-branded gaming handheld later in 2025 through partnership with a PC gaming manufacturer, followed by next-generation Xbox consoles targeting 2027, according to WindowsCentral. The handheld device, codenamed "Keenan," will run full Windows with Microsoft Store and PC Game Pass integration, featuring distinct Xbox design elements including an official guide button.

Microsoft is expected to use the device to test new Windows 11 "device aware" capabilities while reducing typical OEM bloatware. The next-generation Xbox console plans include a premium successor to Xbox Series X alongside Microsoft's own gaming handheld and new controller options with direct-to-cloud connectivity, the report said. These consoles will reportedly operate closer to Windows architecture, reducing developer porting workloads while maintaining backwards compatibility with legacy Xbox games.
Space

That Galaxy Next Door? It's Home to a Monster Black Hole (npr.org) 27

NPR reports on "a monster black hole that's been lurking unseen in the galaxy next door." This appears to be the closest supermassive black hole outside our Milky Way galaxy, according to a report that's appearing in The Astrophysical Journal... "Now that there is strong evidence that it should be there, you can rest assured that we are very excitedly following up," says Jesse Han of the Center for Astrophysics, Harvard & Smithsonian, who led the study...

Han and his colleagues realized that this black hole must exist when they were studying so-called hypervelocity stars... [T]hey started out as normal stars that were part of a binary system, or two stars orbiting each other. When that kind of pair ventures too close to a supermassive black hole, says Han, "what can happen is one of the stars can get captured by the black hole. It is basically ripped apart from its companion." The bereft companion star, meanwhile, gets flung away, going at ridiculously high speeds. It's as if the black hole basically hurled it out of the galaxy.

And that explains some of what's happening in our own galaxy, writes Space.com: Tracing the trajectory of these super-speed stars using the European Space Agency's star-tracking Gaia satellite, the researchers discovered that around half of them were accelerated by the Milky Way's own supermassive black hole, Sagittarius A*. The other half, the team believes, likely fled to the outskirts of the Milky Way after a gravitational encounter with a supermassive black hole at the heart of the LMC separated these stars from their stellar binary partners.

"It is astounding to realize that we have another supermassive black hole just down the block, cosmically speaking," team leader Jesse Han of the Center for Astrophysics | Harvard & Smithsonian (CfA) said in a statement. "Black holes are so stealthy that this one has been practically under our noses this whole time."

"Their calculations suggest that the Large Magellanic Cloud must be harboring a black hole that's about 600,000 times the mass of our Sun," adds NPR. "That's smaller than the black hole at the center of the Milky Way, which is about 4 million times more massive than the Sun." Astronomers had previously thought the Large Magellanic Cloud should have a big black hole, but until now there's been no evidence of it, says Han... Now, though, astronomers have a better sense of where to hunt for any X-ray, radio, or visible-light signatures that are the telltale signs of an invisible black hole that's devouring everything nearby. "It is within the realm of possibilities that it is already detectable in radio and X-ray and optical," says Han. "We just haven't looked at the right place."
AI

Will an 'AI Makeover' Help McDonald's? (msn.com) 100

"McDonald's is giving its 43,000 restaurants a technology makeover," reports the Wall Street Journal, including AI-enabled drive-throughs and AI-powered tools for managers — as well as internet-connected kitchen equipment.

"Technology solutions will alleviate the stress...." says McDonald's CIO Brian Rice. McDonald's tapped Google Cloud in late 2023 to bring more computing power to each of its restaurants — giving them the ability to process and analyze data on-site... a faster, cheaper option than sending data to the cloud, especially in more far-flung locations with less reliable cloud connections, said Rice... Edge computing will enable applications like predicting when kitchen equipment — such as fryers and its notorious McFlurry ice cream machines — is likely to break down, Rice said. The burger chain said its suppliers have begun installing sensors on kitchen equipment that will feed data to the edge computing system and give franchisees a "real-time" view into how their restaurants are operating. AI can then analyze that data for early signs of a maintenance problem.

McDonald's is also exploring the use of computer vision, the form of AI behind facial recognition, in store-mounted cameras to determine whether orders are accurate before they're handed to customers, he said. "If we can proactively address those issues before they occur, that's going to mean smoother operations in the future," Rice added...

Additionally, the ability to tap edge computing will power voice AI at the drive-through, a capability McDonald's is also working with Google's cloud-computing arm to explore, Rice said. The company has been experimenting with voice-activated drive-throughs and robotic deep fryers since 2019, and ended its partnership with International Business Machines to test automated order-taking at the drive-through in 2024.

Edge computing will also help McDonald's restaurant managers oversee their in-store operations. The burger giant is looking to create a "generative AI virtual manager," Rice said, which handles administrative tasks such as shift scheduling on managers' behalf. Fast-food giant Yum Brands' Pizza Hut and Taco Bell have explored similar capabilities.

ISS

Axiom Space and Red Hat Will Bring Edge Computing to the International Space Station (theregister.com) 7

Axiom Space and Red Hat will collaborate to launch Data Center Unit-1 (AxDCU-1) to the International Space Station this spring. It's a small data processing prototype (powered by lightweight, edge-optimized Red Hat Device Edge) that will demonstrate initial Orbital Data Center (ODC) capabilities.

"It all sounds rather grand for something that resembles a glorified shoebox," reports the Register. Axiom Space said: "The prototype will test applications in cloud computing, artificial intelligence, and machine learning (AI/ML), data fusion and space cybersecurity."

Space is an ideal environment for edge devices. Connectivity to datacenters on Earth is severely constrained, so the more processing that can be done before data is transmitted to a terrestrial receiving station, the better. Tony James, chief architect, Science and Space at Red Hat, said: "Off-planet data processing is the next frontier, and edge computing is a crucial component. With Red Hat Device Edge and in collaboration with Axiom Space, Earth-based mission partners will have the capabilities necessary to make real-time decisions in space with greater reliability and consistency...."

The Red Hat Device Edge software used by Axiom's device combines Red Hat Enterprise Linux, the Red Hat Ansible Platform, and MicroShift, a lightweight Kubernetes container orchestration service derived from Red Hat OpenShift. The plan is for Axiom Space to host hybrid cloud applications and cloud-native workloads on-orbit. Jason Aspiotis, global director of in-space data and security, Axiom Space, told The Register that the hardware itself is a commercial off-the-shelf unit designed for operation in harsh environments... "AxDCU-1 will have the ability to be controlled and utilized either via ground-to-space or space-to-space communications links. Our current plans are to maintain this device on the ISS. We plan to utilize this asset for at least two years."

The article notes that HPE has also "sent up a succession of Spaceborne computers — commercial, off-the-shelf supercomputers — over the years to test storage, recovery, and operational potential on long-duration missions." (They apparently use Red Hat Enterprise Linux.) "At the other end of the scale, the European Space Agency has run Raspberry Pi computers on the ISS for years as part of the AstroPi educational outreach program."

Axiom Space says their Orbital Data Center is deigned to "reduce delays traditionally associated with orbital data processing and analysis." By utilizing Earth-independent cloud storage and edge processing infrastructure, Axiom Space ODCs will enable data to be processed closer to its source, spacecraft or satellites, bypassing the need for terrestrial-based data centers. This architecture alleviates reliance on costly, slow, intermittent or contested network connections, creating more secure and quicker decision-making in space.

The goal is to allow Axiom Space and its partners to have access to real-time processing capabilities, laying the foundation for increased reliability and improved space cybersecurity with extensive applications. Use cases for ODCs include but are not limited to supporting Earth observation satellites with in-space and lower latency data storage and processing, AI/ML training on-orbit, multi-factor authentication and cyber intrusion detection and response, supervised autonomy, in-situ space weather analytics and off-planet backup & disaster recovery for critical infrastructure on Earth.

AI

Signal President Calls Out Agentic AI As Having 'Profound' Security and Privacy Issues (techcrunch.com) 8

Signal President Meredith Whittaker warned at SXSW that agentic AI poses significant privacy and security risks, as these AI agents require extensive access to users' personal data, likely processing it unencrypted in the cloud. TechCrunch reports: "So we can just put our brain in a jar because the thing is doing that and we don't have to touch it, right?," Whittaker mused. Then she explained the type of access the AI agent would need to perform these tasks, including access to our web browser and a way to drive it as well as access to our credit card information to pay for tickets, our calendar, and messaging app to send the text to your friends. "It would need to be able to drive that [process] across our entire system with something that looks like root permission, accessing every single one of those databases -- probably in the clear, because there's no model to do that encrypted," Whittaker warned.

"And if we're talking about a sufficiently powerful ... AI model that's powering that, there's no way that's happening on device," she continued. "That's almost certainly being sent to a cloud server where it's being processed and sent back. So there's a profound issue with security and privacy that is haunting this hype around agents, and that is ultimately threatening to break the blood-brain barrier between the application layer and the OS layer by conjoining all of these separate services [and] muddying their data," Whittaker concluded.

If a messaging app like Signal were to integrate with AI agents, it would undermine the privacy of your messages, she said. The agent has to access the app to text your friends and also pull data back to summarize those texts. Her comments followed remarks she made earlier during the panel on how the AI industry had been built on a surveillance model with mass data collection. She said that the "bigger is better AI paradigm" -- meaning the more data, the better -- had potential consequences that she didn't think were good. With agentic AI, Whittaker warned we'd further undermine privacy and security in the name of a "magic genie bot that's going to take care of the exigencies of life," she concluded.
You can watch the full speech on YouTube.
Businesses

70% of Large VMware Customers Bought Broadcom's Biggest Bundle (theregister.com) 30

Broadcom's VMware acquisition has significantly boosted revenue, largely driven by high-priced VMware Cloud Foundation bundles adopted by the majority of its top customers. The Register reports: 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. The chips-and-code company today announced its results for the quarter ended February 2nd, its first for FY 2025. Revenue of $14.92 billion represented 25 percent year-on-year growth. Net income of $5.5 billion was a 315 percent increase on the result from Q1 2024.

Broadcom no longer breaks out VMware revenue: sales of Virtzilla's wares are all now lumped into its infrastructure software business unit, which posted $6.7 billion revenue for Q1, up from $4.55 billion for the same quarter last year. Direct comparisons of those numbers are not wise as Broadcom owned VMware for four fifths of Q1 2024. Consider, instead, the $1.97 billion Q4 2023 and $7.6 billion FY 2023 software revenue that Broadcom recorded before it acquired VMware.

Know, also, that Broadcom's software sales grew by just three percent in FY 2023 and four percent in FY 2022. That slow growth means the jump from $1.97 billion software revenue in Q4 2023 to $6.7 billion in Q1 2025 is likely due to VMware, which in its last quarter as an independent company reported $3.4 billion revenue. It therefore looks a lot like Broadcom has added around $1 billion to quarterly VMware revenue in a little over a year.

AI

US Likely To Ban Chinese App DeepSeek From Government Devices (msn.com) 14

The White House is weighing measures to restrict Chinese artificial-intelligence upstart DeepSeek, including banning its chatbot from government devices because of national-security concerns, WSJ reported Friday, citing people familiar with the matter. From the report: U.S. officials are worried about DeepSeek's handling of user data, which the Chinese company says it stores in servers located in China, the people said. Officials also believe DeepSeek hasn't sufficiently explained how it uses the data it collects and who has access to the data, they said.

The Trump administration is likely to adopt a rule that would bar people from downloading DeepSeek's chatbot app onto U.S. government devices, the people said. Officials are also considering two other possible moves: banning the DeepSeek app from U.S. app stores and putting limits on how U.S.-based cloud service providers could offer DeepSeek's AI models to their customers, people close to the matter said. They cautioned that discussions about these two moves were still at an early stage.

AI

Mistral Adds a New API That Turns Any PDF Document Into an AI-Ready Markdown File 24

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.

Slashdot Top Deals