Google unveiled Gemini 2.0 yesterday, almost exactly one year after Google's initial Gemini launch. The new release offers enhanced multimodal capabilities like native image and audio output, real-time tool use, and advanced reasoning to enable agentic experiences, such as acting as a universal assistant or research companion. VentureBeat reports: During a recent press conference, Tulsee Doshi, director of product management for Gemini, outlined the system's enhanced capabilities while demonstrating real-time image generation and multilingual conversations. "Gemini 2.0 brings enhanced performance and new capabilities like native image and multilingual audio generation," Doshi explained. "It also has native intelligent tool use, which means that it can directly access Google products like search or even execute code."

The initial release centers on Gemini 2.0 Flash, an experimental version that Google claims operates at twice the speed of its predecessor while surpassing the capabilities of more powerful models. This represents a significant technical achievement, as previous speed improvements typically came at the cost of reduced functionality. Perhaps most significantly, Google introduced three prototype AI agents built on Gemini 2.0's architecture that demonstrate the company's vision for AI's future. Project Astra, an updated universal AI assistant, showcased its ability to maintain complex conversations across multiple languages while accessing Google tools and maintaining contextual memory of previous interactions. [...]

For developers and enterprise customers, Google introduced Project Mariner and Jules, two specialized AI agents designed to automate complex technical tasks. Project Mariner, demonstrated as a Chrome extension, achieved an impressive 83.5% success rate on the WebVoyager benchmark for real-world web tasks -- a significant improvement over previous attempts at autonomous web navigation. Supporting these advances is Trillium, Google's sixth-generation Tensor Processing Unit (TPU), which becomes generally available to cloud customers today. The custom AI accelerator represents a massive investment in computational infrastructure, with Google deploying over 100,000 Trillium chips in a single network fabric.

Amazon has postponed implementing Microsoft's cloud-based Office suite for its workforce by one year, citing security concerns following a Russian cyber attack on Microsoft's systems. The delay affects a $1 billion, five-year contract signed last year to provide Microsoft 365 to Amazon's 1.5 million employees, making the e-commerce giant one of the largest customers of Microsoft's cloud productivity suite.

The decision came after Microsoft revealed that Midnight Blizzard, a Russia-linked hacking group, had breached several employee email accounts, including those of senior executives and cybersecurity staff. Amazon subsequently conducted its own security review and requested enhanced protection measures from Microsoft.

Google today unveiled Project Mariner, its first AI agent capable of autonomously navigating web browsers, operating through a Chrome extension that controls cursor movements and form-filling to replicate human interactions online.

The Gemini-powered prototype, developed by Google's DeepMind division, is initially available to a select group of testers. During demos, the agent performed tasks like creating shopping carts on grocery websites, though with noticeable five-second delays between actions. The system captures browser screenshots and processes them through Gemini in the cloud to generate navigation commands.

It operates only in Chrome's active tab, requiring users to observe its actions rather than running in the background. Project Mariner achieved an 83.5% success rate on the WebVoyager benchmark for web-based tasks. The agent has built-in limitations, including inability to complete purchases, accept cookies, or agree to terms of service. Google Labs Director Jaclyn Konzelmann described the project as a "fundamentally new UX paradigm shift" that could transform how users interact with websites. The company said it is engaging with web ecosystem stakeholders as development continues.

An anonymous reader shares a report: Google recently asked the U.S. government to break up Microsoft's exclusive agreement to host OpenAI's technology on its cloud servers, according to a person who has been directly involved in the effort. The conversation took place after the Federal Trade Commission, one of the primary federal antitrust enforcement agencies, asked Google about Microsoft's business practices as part of a broader investigation, this person said.

Firms that compete with Microsoft in renting out cloud servers, including Google and Amazon, want to host OpenAI's artificial intelligence themselves so their cloud customers don't need to also tap Microsoft servers to get access to the startup's technology, this person said.

This year Marvell's stock rose 95%, giving it a $100 billion market capitalization, according to the Wall Street Journal.

"The latest gains have even put Marvell's market cap ahead of much-beleaguered Intel, which still generates 10 times as much annual revenue." Marvell's recent trajectory suggests that the revenue gap will continue to narrow. The explosive growth of its data center business has finally reached a point where it can more fully offset weakness in the company's more legacy segments, which sell chips used in goods such as telecommunications gear, cable TV boxes and autos. Data center sales nearly doubled year over year to $1.1 billion in the just-ended quarter, and Marvell's projection for the current period indicates the company will end its fiscal year in January with the data center unit encompassing about 72% of its total revenue, up from 40% in the previous year.

The next year is looking bright as well. Marvell's latest deal with Amazon is a five-year "multigenerational" agreement that has Marvell helping Amazon design its own artificial intelligence chips. Amazon, which runs the world's largest cloud computing service, has been expanding its internal chip efforts significantly, in part to reduce its reliance on Nvidia for crucial AI components. Amazon announced the next generation of its largest AI chip, called Trainium, at its annual developers conference this week. Analysts believe Trainium will play a role in Marvell's AI custom revenue more than doubling in the next fiscal year ending January of 2026.

That is expected to help propel Marvell's annual revenue to more than $8 billion in fiscal 2026, up 40% from what is expected for this year, according to consensus estimates from Visible Alpha. In addition, 20% growth is expected for the following year, when Marvell expects to be in production of custom AI chips for another unnamed big tech customer that analysts believe to be Microsoft. Analyst Mark Lipacis of Evercore ISI projects that the industry for custom AI chips will reach $30 billion to $50 billion in sales by 2030. In a note to clients last week, he said Marvell "has the potential to capture one-third of that market."
Marvell's CEO "has been among the few names floated as potential replacements for the recently ousted Pat Gelsinger at Intel's corner office," the article points out — which meant he had to reassure investors on an earnings call that he was staying at Marvell.

"The company is outstanding. The technology is best-in-class. I can't think of a better place to work than Marvell."

This week AWS pledged up to $100 million in cloud-computing credits for educational organizations over the next five years, to help them build "technology-based learning experiences" on AWS, including:

• AI assistants
• coding curriculums - connectivity tools
• student learning platforms
• mobile apps
• chatbots

One example shared by Amazon: The nonprofit Code.org will use AWS's cloud credits to scale their AI teaching assistant that "has already helped teachers reduce the time they spend assessing students' coding projects by up to 50%." (Amazon's blog post notes that "Improved efficiency means teachers have more time to work on personalized lesson plans and coach students" — and that Code.org's assistant uses an AWS service for building AI tools...)

$100 million sounds pretty generous. But long-time Slashdot reader theodp notes the application for the cloud credits limits education organization to $100,000 in credits (though "your organization may be able to apply for a credits expansion" if needed). Do these figures suggest Amazon expects less than 1,000 organizations to apply for free cloud-computing over the next five years? ($100,000,000/$100,000 = 1,000)

theodp also spotted a GitHub comment from a Code.org software engineer comparing accuracy for its teaching assistant after a switch from GPT-4 Turbo to Claude. Both before and after the switch, the teaching assistant averaged an accuracy rate of 77%, the comment notes.

I guess that 77% accuracy rate is what Amazon is calling "improved efficiency" that "means teachers have more time to work on personalized lesson plans and coach students." (Maybe you're never to young to learn that AI makes mistakes?)

Today a group of German scientists presented data suggesting Earth is absorbing more sunlight than in the past, reports Ars Technica, "largely due to reduced cloud cover." We can measure both the amount of energy the Earth receives from the Sun and how much energy it radiates back into space.... The new paper finds that the energy imbalance set a new high in 2023, with a record amount of energy being absorbed by the ocean/atmosphere system. This wasn't accompanied by a drop in infrared emissions from the Earth, suggesting it wasn't due to greenhouse gases, which trap heat by absorbing this radiation. Instead, it seems to be due to decreased reflection of incoming sunlight by the Earth....

Using two different data sets, the teams identify the areas most effected by this, and they're not at the poles, indicating loss of snow and ice are unlikely to be the cause. Instead, the key contributor appears to be the loss of low-level clouds [particularly over the Atlantic ocean]... The drop in low-level clouds had been averaging about 1.3 percent per decade. 2023 saw a slightly larger drop occur in just one year....

So, what could be causing the clouds to go away? The researchers list three potential factors. One is simply the variability of the climate system, meaning 2023 might have just been an extremely unusual year, and things will revert to trends in the ensuing years. The second is the impact of aerosols, which both we and natural processes emit in copious quantities. These can help seed clouds, so a reduction of aerosols (driven by things like pollution control measures) could potentially account for this effect. The most concerning potential explanation, however, is that there may be a feedback relationship between rising temperatures and low-level clouds. Meaning that, as the Earth warms, the clouds become sparse, enhancing the warming further. That would be bad news for our future climate, because it suggests that the lower range of warming estimates would have to be adjusted upward to account for it.
If the decline in reflectivity wasn't just caused by normal variability, the researchers warn, "the 2023 extra heat may be here to stay..."

An anonymous reader shares a report: Companies have been discussing migrating off of VMware since Broadcom's takeover a year ago led to higher costs and other controversial changes. Now we have an inside look at one of the larger customers that recently made the move.

According to a report from The Register today, Beeks Group, a cloud operator headquartered in the United Kingdom, has moved most of its 20,000-plus virtual machines (VMs) off VMware and to OpenNebula, an open source cloud and edge computing platform. Beeks Group sells virtual private servers and bare metal servers to financial service providers. It still has some VMware VMs, but "the majority" of its machines are currently on OpenNebula, The Register reported.

Beeks' head of production management, Matthew Cretney, said that one of the reasons for Beeks migration was a VMware bill for "10 times the sum it previously paid for software licenses," per The Register. According to Beeks, OpenNebula has enabled the company to dedicate more of its 3,000 bare metal server fleet to client loads instead of to VM management, as it had to with VMware. With OpenNebula purportedly requiring less management overhead, Beeks is reporting a 200 percent increase in VM efficiency since it now has more VMs on each server.

Intel CEO Pat Gelsinger has stepped down amid the company's continued struggles against rivals, with shares losing over half their value this year. The chipmaker announced Monday that Chief Financial Officer David Zinsner and Executive Vice President Michelle Johnston Holthaus will serve as interim co-CEOs while the board searches for a permanent replacement.

Gelsinger, 63, was hired in 2021 to lead an ambitious turnaround aimed at reclaiming Intel's technological edge from competitors like Taiwan Semiconductor Manufacturing Co. His strategy included expanding Intel's factory network with new facilities in Ohio and transforming the company into a contract manufacturer for other firms. The plan faced significant headwinds as Nvidia dominated the AI chip market, with cloud computing companies increasingly favoring Nvidia's processors for AI development over Intel's Gaudi line.

Intel's challenges culminated in an August earnings report showing a surprise loss, leading to dividend suspension and plans to cut over 15% of its 110,000-person workforce. Board Chairman Frank Yeary, now serving as interim executive chair, emphasized the need to prioritize Intel's product group to meet customer demands. The leadership change also impacts the Biden administration's semiconductor industry initiatives, as Intel was set to receive the largest grant under the $39 billion Chips Act program.

Multiple news outlets including Bloomberg and Reuters report that Gelsinger was forced out by the board because "directors felt Gelsinger's costly and ambitious plan to turn Intel around was not working and the progress of change was not fast enough."

An anonymous reader shares a report: Google Cloud dangled hundreds of million of euros worth of financial incentives to ally itself with an association of European cloud providers that had lodged a complaint against Microsoft, according to confidential documents seen by The Register.

Amit Zavery, the former Vice President of Google Cloud Platform, presented to a selection of members of the Cloud Infrastructure Service Providers in Europe (CISPE) trade body, then to the board and finally to the entire organization, according to sources that asked to remain anonymous.

In the presentation, seen by us, Zavery offered to provide a Members Innovation Fund of $4.2 million, which Google described as $105,000 per member to be used as "immediate funding for projects and license fees of CISPE members to support innovation in open cloud ecosystems." CISPE actually has 36 members now, including Oxya, Leaseweb, UpCloud and AWS -- the latter being the only non-European participant. The number has grown from 27 in July. Google also offered to contribute an additional $10.6 million to the trade association, described in the presentation as "participating and membership resources."

The FTC has opened a broad antitrust investigation into Microsoft, including of its software licensing and cloud computing business. Bloomberg first reported the news. Reuters reports: The probe was approved by FTC Chair Lina Khan ahead of her likely departure in January. The election of Donald Trump as U.S. president and the expectation he will appoint a fellow Republican with a softer approach toward business, leaves the outcome of the investigation up in the air.

The FTC is examining allegations that the software giant is potentially abusing its market power in productivity software by imposing punitive licensing terms to prevent customers from moving their data from its Azure cloud service to other competitive platforms, sources confirmed earlier this month. The FTC is also looking at practices related to cybersecurity and artificial intelligence products, the source said on Wednesday.

The Verge's Wes Davis reports: A new startup created by former Android leaders aims to build an operating system for AI agents. Among them is Hugo Barra, Google's former VP of Android product management, who says the new company -- named "/dev/agents" -- will revisit the leaders' "Android roots."

"We can see the promise of AI agents, but as a developer, it's just too hard to build anything good," /dev/agents cofounder and CEO and Google's former Android VP of engineering David Singleton told Bloomberg. He said the industry needs "an Android-like moment for AI."

The company is working on a cloud-based "next-gen operating system for AI agents" intended "for trusted agents to work with users across all of their devices," Singleton wrote in a post on X. He said that AI agents will "need new UI patterns, a reimagined privacy model, and a developer platform that makes it radically simpler to build useful agents."

An anonymous reader quotes a report from KrebsOnSecurity: Two men have been arrested for allegedly stealing data from and extorting dozens of companies that used the cloud data storage company Snowflake, but a third suspect -- a prolific hacker known as Kiberphant0m -- remains at large and continues to publicly extort victims. However, this person's identity may not remain a secret for long: A careful review of Kiberphant0m's daily chats across multiple cybercrime personas suggests they are a U.S. Army soldier who is or was recently stationed in South Korea.

Kiberphant0m's identities on cybercrime forums and on Telegram and Discord chat channels have been selling data stolen from customers of the cloud data storage company Snowflake. At the end of 2023, malicious hackers discovered that many companies had uploaded huge volumes of sensitive customer data to Snowflake accounts that were protected with nothing more than a username and password (no multi-factor authentication required). After scouring darknet markets for stolen Snowflake account credentials, the hackers began raiding the data storage repositories for some of the world's largest corporations. Among those was AT&T, which disclosed in July that cybercriminals had stolen personal information, phone and text message records for roughly 110 million people. Wired.com reported in July that AT&T paid a hacker $370,000 to delete stolen phone records.

On October 30, Canadian authorities arrested Alexander Moucka, a.k.a. Connor Riley Moucka of Kitchener, Ontario, on a provisional arrest warrant from the United States, which has since indicted him on 20 criminal counts connected to the Snowflake breaches. Another suspect in the Snowflake hacks, John Erin Binns, is an American who is currently incarcerated in Turkey. Investigators say Moucka, who went by the handles Judische and Waifu, had tasked Kiberphant0m with selling data stolen from Snowflake customers who refused to pay a ransom to have their information deleted. Immediately after news broke of Moucka's arrest, Kiberphant0m was clearly furious, and posted on the hacker community BreachForums what they claimed were the AT&T call logs for President-elect Donald J. Trump and for Vice President Kamala Harris. [...] Also on Nov. 5, Kiberphant0m offered call logs stolen from Verizon's push-to-talk (PTT) customers -- mainly U.S. government agencies and emergency first responders. Kiberphant0m denies being in the U.S. Army and said all these clues were "a lengthy ruse designed to create a fictitious persona," reports Krebs.

"I literally can't get caught," Kiberphant0m said, declining an invitation to explain why. "I don't even live in the USA Mr. Krebs." A mind map illustrates some of the connections between and among Kiberphant0m's apparent alter egos.

Blue Yonder, a Panasonic subsidiary specializing in AI-driven supply chain solutions, experienced a recent ransomware attack that impacted many of its customers. "Among its 3,000 customers are high-profile organizations like DHL, Renault, Bayer, Morrisons, Nestle, 3M, Tesco, Starbucks, Ace Hardware, Procter & Gamble, Sainsbury, and 7-Eleven," reports BleepingComputer. From the report: On Friday, the company warned that it was experiencing disruptions to its managed services hosting environment due to a ransomware incident that occurred the day before, on November 21. "On November 21, 2024, Blue Yonder experienced disruptions to its managed services hosted environment, which was determined to be the result of a ransomware incident," reads the announcement. "Since learning of the incident, the Blue Yonder team has been working diligently together with external cybersecurity firms to make progress in their recovery process. We have implemented several defensive and forensic protocols."

Blue Yonder claims it has detected no suspicious activity in its public cloud environment and is still processing multiple recovery strategies. [...] As expected, this has impacted clients directly, as a spokesperson for UK grocery store chain Morrisons has confirmed to the media they have reverted to a slower backup process. Sainsbury told CNN that it had contingency plans in place to overcome the disruption. A Saturday update informed customers that the restoration of the impacted services continued, but no specific timelines for complete restoration could be shared yet. Another update published on Sunday reiterated the same, urging clients to monitor the customer update page on Blue Yonder's website over the coming days.

An anonymous reader quotes a report from ZDNet, written by Steven Vaughan-Nichols: At KubeCon North America, SUSE announced a significant rebranding effort, several new product offerings, and the launch of SUSE AI, a secure platform for deploying and running generative AI (gen AI) applications. SUSE has renamed its entire portfolio to make product names more descriptive and customer-friendly. Notable changes include:

- Rancher, SUSE's Kubernetes offering, is now SUSE Rancher.
- Liberty Linux, the company's Red Hat Enterprise Linux (RHEL)/CentOS clone and support offering, becomes SUSE Multi Linux Support.
- Harvester is rebranded as SUSE Virtualization
- Longhorn is now SUSE Storage.

[...] Also, like everyone else, SUSE now has an AI offering: SUSE AI. This isn't an AI chatbot, like Red Hat's Lightspeed AI tool. No, it's a secure platform for deploying and running gen AI applications. This new offering addresses key challenges faced by enterprises as they move from AI experimentation to deployment, particularly in areas of security and compliance. These are SUSE AI's top features, as highlighted by Vaughan-Nichols:

1. Security by Design: SUSE AI provides security and certifications at the software infrastructure level, along with zero-trust security tools, templates, and compliance playbooks.
2. Multifaceted Trust: The platform ensures that generated data is correct and private customer and IP data remain secure. It supports deployment across various environments, including on-premise, hybrid, cloud, and air-gapped setups.
3. Choice and Flexibility: SUSE AI allows customers to select and deploy their preferred AI components and LLMs.
4. Simplified Operations: The platform provides simplified cluster operations, persistent storage, and easy access to pre-configured shared tools and services.

Google is introducing "Restore Credentials," a feature that simplifies transferring app credentials when switching Android devices to keep you logged into your apps. The Verge reports: While some apps already did this, Google is making it easier for developers to include this experience by implementing a "restore key" that automatically transfers to the new phone and logs you back into the app. [...] Restore Credentials requires less work than the previous approach on Android, and can automatically check if a restore key is available and log you back in at the first app launch. A restore key is a public key that uses existing passkey infrastructure to move about your credentials.

Restore keys can also be backed up to the cloud, although developers can opt out. For that reason, transferring directly from device to device will still likely be more thorough than restoring from the cloud, as is the case with Apple devices today. Notably, Google says restore keys do not transfer if you delete an app and reinstall it.

ZDNet's Steven Vaughan-Nichols interviews Jim Zemlin, Executive Director of The Linux Foundation and "head janitor of open source." An anonymous Slashdot reader shares an excerpt from the article: When I first met Zemlin, he was the head of the Free Standards Group (FSG). The FSG's main project was the Linux Standard Base (LSB) project. The LSB's goal was to get everyone in the Linux desktop world to agree on standards to ensure compatibility among distributions and their applications. Oh well, some struggles are never-ending. Another group, the Open Source Development Labs (OSDL), was simultaneously working on standardizing enterprise Linux. The two non-profits had the same goal of making Linux more useful and popular, so they agreed to merge. Zemlin was the natural pick to head this new group, which would be called The Linux Foundation.

At the time, he told me: "The combination of the two groups really enables the Linux platform and all the members of the Linux Foundation to work really effectively. I clearly understand what the organization's charter needs to be: We need to provide services that are useful to the community and industry, as well as protect, promote, and continue to standardize the platform." While initially focused on Linux, the Foundation's scope expanded significantly around 2010. Until then, the organization had hosted about a dozen projects related to the Linux operating system. However, as Linux gained dominance in various sectors, including high-performance computing, automotive, embedded systems, mobile devices, and cloud computing, the Linux Foundation started to broaden its horizons. Zemlin says there are three words that sum up the Linux Foundation's effort to keep open source safe and open to a new generation of developers: helpful, hopeful, and humble.

"You must be genuinely helpful to developers. We're the janitors of open source. The Linux Foundation takes care of all the boring but important stuff necessary to support software development so developers can focus on code. This work includes events, project marketing, project infrastructure, finances for projects, training and education, legal assistance, standards, facilitation, open source evangelism, and much, much more."

He continued: "The hopeful part is really the optimistic part. When in 2007, people were saying that this would never work. When leaders of huge companies tell everyone that you know all that you're doing is a cancer or terrible, you have to have a sense of optimism that there are better days ahead. You have to always be thinking, 'No, we can do it and stick with it.'"

However, Zemlin concluded that the number one trait that's "important in working in open source is this idea of humility. I work with hundreds of people every day, and none of them work at the Linux Foundation. We must lead through influence, and that really has been the secret for 20 years of working here without going totally insane. If you can check your ego and take criticism, open source actually turns out to be a really fun community to work with."

An anonymous reader quotes a report from GeekWire: Microsoft and Atom Computing say they've reached a new milestone in their effort to build fault-tolerant quantum computers that can show an advantage over classical computers. Microsoft says it will start delivering the computers' quantum capabilities to customers by the end of 2025, with availability via the Azure cloud service as well as through on-premises hardware. "Together, we are co-designing and building what we believe will be the world's most powerful quantum machine," Jason Zander, executive vice president at Microsoft, said in a LinkedIn posting.

Like other players in the field, Microsoft's Azure Quantum team and Atom Computing aim to capitalize on the properties of quantum systems -- where quantum bits, also known as qubits, can process multiple values simultaneously. That's in contrast to classical systems, which typically process ones and zeros to solve algorithms. Microsoft has been working with Colorado-based Atom Computing on hardware that uses the nuclear spin properties of neutral ytterbium atoms to run quantum calculations. One of the big challenges is to create a system that can correct the errors that turn up during the calculations due to quantum noise. The solution typically involves knitting together "physical qubits" to produce an array of "logical qubits" that can correct themselves.

In a paper posted to the ArXiv preprint server, members of the research team say they were able to connect 256 noisy neutral-atom qubits using Microsoft's qubit-virtualization system in such a way as to produce a system with 24 logical qubits. "This represents the highest number of entangled logical qubits on record," study co-author Krysta Svore, vice president of advanced quantum development for Microsoft Azure Quantum, said today in a blog posting. "Entanglement of the qubits is evidenced by their error rates being significantly below the 50% threshold for entanglement." Twenty of the system's logical qubits were used to perform successful computations based on the Bernstein-Vazirani algorithm, which is used as a benchmark for quantum calculations. "The logical qubits were able to produce a more accurate solution than the corresponding computation based on physical qubits," Svore said. "The ability to compute while detecting and correcting errors is a critical component to scaling to achieve scientific quantum advantage."

Sony is bringing cloud streaming to the PlayStation Portal. "When it first launched, the device was only able to stream games from your PS5 over Wi-Fi," notes The Verge's Jay Peters. "But as part of a new system update that's rolling out starting later today, you'll be able to stream select PS5 games from the PlayStation Plus Game Catalog to your PlayStation Portal." From the report: Sony is launching the feature in beta, and you'll need to be a PlayStation Plus Premium subscriber to take advantage of it. Sony says that to stream at 720p, you'll need a minimum 7 Mbps connection, while 1080p quality will require a minimum 13 Mbps connection. Some PlayStation Plus features won't be available to start with cloud streaming to the PlayStation Portal, including Game Trials, party voice chat, game invites for select games, 3D audio, and "in-game commerce." And you won't be able to stream any PS4 games or PS3 games. Child accounts also won't be able to use cloud streaming on the Portal.

Microsoft is planning to launch a new purpose-built miniature PC for its Windows 365 cloud service next year. The Verge: Windows 365 Link is a $349 device that acts like a thin client PC to connect to the cloud and stream a version of Windows 11. The Link device is designed to be a compact, fanless, and easy-to-use cloud PC for your local monitors and peripherals. It's meant to be the ideal companion to Microsoft's Windows 365 service, which lets businesses transition employees over to virtual machines that exist in the cloud and can be streamed securely to multiple devices. Windows 365 Link cannot run local apps.