News: Technology News

Weather from Microsoft Start is once again named as the most accurate world forecast provider

Wed, 29 May 2024 02:12:51 +0000

Amidst all of the attention that has been given to Microsoft's generative AI features, and also its various gaming divisions, another division at the company has been given special recognition by an assessment group for the second year in a row.

In a blog post, Microsoft revealed that its weather forecasting team at its Microsoft Start division has been named as the "overall most accurate weather forecast provider globally" by Forecastwatch for 2023. This is the same assessment company that awarded the Microsoft Start division the same prize a year ago for its forecasting methods in 2022.

The blog post states:

This independent study of 22 weather services for the entire year of 2023 evaluated 13 different metrics. Weather from Microsoft Start had the top scores for multiple metrics including high and low temperature predictions and cloud cover forecasts. This included forecasts for 1-5 days out. In fact, Weather from Microsoft Start’s total first place percentages were 11% higher than last year and the lead from the nearest competitor grew by 4%.

Weather from Microsoft Start has used AI to help combine weather data with "large scale computation models" along with mapping and good old-fashioned reports from real humans.

In just the past few weeks, Microsoft has revealed that the Start team has been developing all new AI models to help predict weather more accurately from as long as 30 days from now. It has also created new models using both radar and satellite data to better predict when rain and cloud cover might happen.

Microsoft added that just in March 2024 it sent 150 billion weather notifications to its users. Nine billion of those notifications were for severe weather alerts Weather from Microsoft Start is available on Windows, along with its Edge web browser, MSN, Bing, and the Microsoft Start mobile app, among other sources.

Source

Microsoft Flight Simulator City Update 7 brings more upgrades to European cities

Tue, 28 May 2024 18:15:58 +0000

Microsoft's Flight Simulator development team has just shipped another free content update for all players, and it is bringing more upgraded vistas for players to view and enjoy while flying. City Update 7 is going back to Europe, arriving as the second part of the December 2023 update for the sim. Watch the trailer above to get a peak at what's included.

The developer today announced that the visual upgrades are hitting six European cities specifically "to captivate and inspire" its audience. This means Sweden’s Stockholm, France's Nice, Monaco, Portugal's Porto, as well as Barcelona and Madrid in Spain are the locales being touched up with this update.

"The Microsoft Flight Simulator team and its partners captured these renowned cities in exquisite detail using recent high resolution aerial imagery, digital elevation data, and TIN (triangulated irregular network) surface modeling," explains the announcement blog post. "This latest release invites simmers to choose their preferred aircraft to explore these vibrant locales, from the French Riviera to the skyscrapers of Stockholm."

Microsoft Flight Simulator City Update VII: European Cities II is now available as an optional download across PC, Xbox Series X|S, Xbox and PC Game Pass, as well as Xbox Cloud Gaming via Game Pass Ultimate. It can be found in the in-game marketplace for free. Players will need to update their game to at least version 1.37.18.0 to jump in.

At the same time, the SC.7 Skyvan aircraft has been added to the game as part of its Local Legend packs. The cargo plane, plus 16 liveries, are available for purchase for $14.99 via the marketplace.

Don't forget that Microsoft is currently also developing a sequel for the popular sim entry. Microsoft Flight Simulator 2024 is coming out sometime this year, and it should be one of the games that appear during the Xbox Games Showcase in June. The upcoming title will bring new mission types like aerial firefighting, air ambulance, and mountain rescue, as well as seasons, storms, and tornadoes, among other features.

Source

Weird YouTube behavior skips some videos to the end when using an ad blocker

Tue, 28 May 2024 07:43:51 +0000

Several users have reported that their YouTube videos automatically skip to the end. And yes, it is a bug instead of a feature YouTube experimented with earlier this month. The weird behavior appears to happen only to those who have installed an ad blocker on their device, according to user reports spotted by 9to5Google.

As per user reports, the video skips to the end when it's played from the beginning and also when the video is replayed. Sometimes, the video starts loading endlessly when the user skips to any part of the video. However, everything starts working as expected when the ad blocker is disabled.

The publication notes that the behavior is mostly visible in the case of users who have installed AdBlock on their devices. However, a very small number of users who didn't have an ad blocker installed reported the behavior.

There is a common belief among many users that this is some new trick YouTube has played to limit people from using ad blockers. However, it could be possible that the video-sharing platform may not be at fault. YouTube was previously accused of slowing down videos and causing crashes, which turned out to be a bug affecting AdBlock and AdBlock Plus.

This doesn't mean the Google-owned company doesn't do anything to stop people from using ad blockers on its platform. YouTube has previously resorted to displaying pop-up messages with countdown timers and issuing warnings that people who use third-party apps that block YouTube ads could see buffering when streaming videos.

YouTube uses various formats of advertisements as a source of revenue and to fund creators on its platform. While there are users who can go to lengths to dodge ads altogether, YouTube's decision to put unskippable ads has been criticized by many.

The company offers the YouTube Premium subscription to remove ads from its videos, better 1080p quality, and let users download videos for offline access. After the latest price hike, it starts at $13.99/mo for the individual plans and goes up to $22.9/mo for the family plans, which can be shared among six users.

Image via Pixabay

Source

You're welcome

Another surprise could be awaiting with AMD Ryzen 9000 in the form of new X870 AM5 boards

Mon, 27 May 2024 21:25:10 +0000

It looks like AMD's upcoming Ryzen 9000 series processors, powered by Zen 5 architecture, will be full of quite a few surprises. The biggest one of them is the apparent lack of support the chips have for Windows 10 according to a supposed Lenovo manager.

The Ryzen 9000 naming itself is a bit surprising considering the 8000 series desktop CPU nomenclature is now held solely by the G-series APUs only with much of it going to mobile SKUs. That's not all though as AMD is seemingly looking to rename Strix Point APUs from 8050 series to "Ryzen AI HX" potentially looking to capitalize on the AI hype and buzz surrounding almost everything nowadays. If you recall, AMD made a big deal about its Ryzen 8000G APUs as they had the "world's first NPU in desktop processors."

The third surprise is allegedly heading our way at Computex 2024 as AMD will be unveiling new 800 series chipset boards that will socket the Ryzen 9000 series processors. The rumour has surfaced from Chinese news outlet Benchlife which has in the past provided credible leaks and it claims it has to do with matching with Intel's chipset naming.

In a news article surrounding a newly released Gigabyte B650E motherboard, the site has added that 800 series board for upcoming 9000 series CPUs "has been confirmed". It writes (Google-translated to English):

GIGABYTE is clearly ready for AMD's next generation Ryzen processors.

Obviously, we can expect various motherboard manufacturers to display AMD and Intel 800 series chip motherboards at Computex 2024; it has been confirmed that the chips used with AMD Ryzen 9000 will be called the 800 series, the same as Intel.

If this turns out to be true, it will mean AMD is going to skip a digit for the first time in its chipset naming. The company started the AM5 (LGA1718) socket with its 600 series boards and it had stuck with this naming since 300-series back when AM4 was launched alongside Zen 1.

Speaking of the 300-series, it is believed AMD started with that number as it wanted to one-up Intel since at that time, Team Blue had 200 series chipsets. So, the rumour certainly seems plausible.

One thing that is not surprising about this news is the timing as we have already had evidence about upcoming AMD boards from chipset driver support that was added recently in April.

Source

You're welcome

xAI just raised a whopping $6 billion to help take on OpenAI, Google and Microsoft

Mon, 27 May 2024 21:24:11 +0000

The interest in funding AI startups continues unabated. Today, xAI, the generative AI company created by former OpenAI founder and later critic Elon Musk, announced it has raised an impressive $6 billion in a Series B funding round.

In a blog post, the company stated some of the people and groups that put money into this funding round included "g Valor Equity Partners, Vy Capital, Andreessen Horowitz, Sequoia Capital, Fidelity Management & Research Company, Prince Alwaleed Bin Talal and Kingdom Holding."

Musk, who is also the owner of Tesla, SpaceX, and X (formerly Twitter), left OpenAI in 2018, a few years after it was founded. He has since been publicly critical of OpenAI, and in March, he filed a lawsuit against the company, claiming it had breached its contract to be a non-profit company.

Musk first announced xAI in July 2023. He launched the first version of its AI chatbot, Grok, in November 2023 to a limited number of users. Grok is now available for X users if they sign up for a premium plan. In April, XAI released the latest version of the chatbot, Grok-1.5V, which added vision features in addition to text prompt support.

Today's blog post stated that xAI has "multiple exciting technology updates and products" in the works that will be revealed in the near future. It added:

The funds from the round will be used to take xAI’s first products to market, build advanced infrastructure, and accelerate the research and development of future technologies.

A recent, but so far unconfirmed, report claims that xAI is working on the next version of Grok and that it is developing a new supercomputer to power its functions. The plan is for the supercomputer to become operational by the end of 2025, and it will need clusters of Nvidia H100 GPUs that will be four times the size of similar AI GPU clusters that are available today.

Source

You're welcome

Windows 11 AI-powered Recall feature can run on unsupported laptops but suffers HUGE performance loss without Copilot+ PC NPUs

Mon, 27 May 2024 21:17:05 +0000

Turns out you can run Windows Recall in Windows 11 without a Copilot+ PC and a powerful NPU, but is it worth it?

What you need to know

Microsoft unveiled a host of next-gen AI features that will ship with Windows 11 exclusively to Copilot+ PCs with powerful NPUs.
A user bypassed these stringent requirements and got Windows Recall to work on Arm64 hardware.
The feature's user experience is laggy and slow, as seen in the video demo, demonstrating the need for a powerful NPU.

While Microsoft's stringent requirements limit its new and controversial Windows 11 Recall feature to Copilot+ PCs with powerful NPUs like Qualcomm's Snapdragon X chips, Windows sleuth Albacore has seemingly found a cheat code that lets the neat feature run on traditional Windows 11 PCs, spotting Arm64 hardware (via TechRadar).

Per the video by the well-known Windows hacker Albacore, the feature functions precisely as Microsoft explained. However, the demo is a tad laggy, reiterating Microsoft's NPU threshold for running AI features in Windows 11 smoothly. As you may know, an NPU (Neural Processing Unit) is designed to help light the processing power load from the GPU and CPU on PCs, fostering better performance.

The video highlights how the screenray function will work on Windows 11 Recall. Screenray is when you enter context-sensitive mode and leverage Windows 11 Recall's capabilities to trace back your steps.

Albacore added that the feature "should theoretically work on Intel/AMD too; OEMs only received Arm64 specific ML model bundles, so there's not much I can do yet."

Windows Recall causes security and privacy concerns

"One of the dreams we've always had is how do we introduce memory," stated Microsoft CEO Satya Nadella while talking to The Wall Street Journal's Joanna Stern about Windows 11's new and controversial AI feature, Windows Recall. "Right? Photographic memory into what you do on your PC? And now we have it."

During Microsoft's special Windows and Surface event, the company announced a host of next-gen AI features shipping to Windows 11 as part of the Windows 11 2024 Update (version 24H2), including Live Captions, Windows Recall, and more. The latter stole the show, raising user concerns over security and privacy issues.

Windows Recall runs in the background of your PC and captures snapshots of everything you see and do, enabling the ability to search for anything you've ever done on your PC with natural language.

Microsoft promises that the feature is 100% privacy-focused since it runs on-device via NPU. It's presented to users as an opt-in experience, and users can disable it at any time via the settings.

What's more, you can restrict the feature from grabbing screenshots from specific apps or websites (which are stored locally on your PC). Users can also choose how long they'd like the screenshots stored and how much space is dedicated to this function.

Interestingly, the only way to get the new Recall feature is by buying a new Copilot+ PC. This is because the feature relies on a powerful NPU for its function. Microsoft has remained quiet if/when it plans to ship the next-gen AI features to traditional Windows 11 PCs without/with less powerful NPUs.

Windows 11 Recall is impressive and controversial, but at the same time, there's so much it can help you achieve. And despite Microsoft's 100% privacy-focused promise and running the feature locally with on-device NPU, users have blatantly expressed their reservations toward the feature.

Some users have even indicated that they'll soon transition to other operating systems like Linux as Windows 10's end-of-support looms. Elon Musk even compares the AI feature to a Black Mirror episode. In the interim, the UK data watchdog is investigating Microsoft's new controversial AI feature to determine the safeguards to protect user privacy.

Source

You're welcome

... and then GPU prices went back up again [Video]

Sun, 26 May 2024 19:33:37 +0000

Paul's Hardware (1.48M subscribers)

May 26, 2024

Video length: 9m 47s

0:00 Welcome to Paul’s Tech News - May 26, 2024

1:50 US to Resume China GPU Tariffs; Container crunch sends freight rates soaring

3:24 RTX 5090 Might Have 16 GDDR7 Modules, “Cinderblock” Cooler from 4090 Ti

4:55 Next-Gen LPDDR6 memory to hit 14.4 Gbps, DDR6 up to 17.6 Gbps

6:04 CAMM2 Standard for Memory is coming to Desktop PCs

TECH BRIEFS

6:56 ASUS ROG Ally X specs: 24GB LPDDR5X-7500, big 80Wh battery

7:41 Cooler Master introduces colored ‘AI Thermal Paste’

8:38 Tesla's gaming computer no longer capable of playing Steam games

Source

You're welcome.

Microsoft Weekly: new Surface devices, a lot of Build announcements and a big Copilot outage

Sun, 26 May 2024 19:00:13 +0000

In this episode of Microsoft Weekly, we look at Microsoft's May 20 event and the latest Surface computers, Build 2024 announcements, Windows 11 24H2 arriving in the Release Preview Channel, app updates, and a lot more.

Table of contents:

A big Microsoft event and Build 2024
Windows 10 and 11 news
Windows Insider Program
Updates are available
Reviews are in
Gaming news
Great deals to check
Random fact about Microsoft

A big Microsoft event and Build 2024

Microsoft's big May 20 event is this week's highlight and the most important part. During the presentation, Microsoft announced "Copilot+ PCs," a series of Windows 11 devices with dedicated Neural Processing Units for next-gen AI capabilities. Shortly after that, Microsoft's partners revealed their first "Copilot+ PCs." There are now models from Dell, Lenovo, HP, Acer, ASUS, etc.

Qualcomm understands that its new PC chips need support from software developers. After all, what's the good of new chips if apps are not optimized for ARM? To address that, Microsoft and Qualcomm partnered to release a Snapdragon Dev Kit for Windows. This small form factor PC has a 12-core processor, 32GB of LPDDR5X memory, and 512GB of storage. No info about the price so far.

But it's not just about apps—the new chips can also run some PC games, and you can check what titles are compatible with them thanks to a newly launched website.

Microsoft itself brought new Surface devices to the stage, revealing the new Surface Pro and Surface Laptop with ARM processors. As expected, these computers feature big display updates, powerful processors from Qualcomm, and other improvements. The new Surface Pro also got a new Surface Pro Flex keyboard, which now has quite a price tag.

If you want to learn more, check out our Specs Appeal articles for the Surface Pro and the Surface Laptop.

One of the most intriguing parts of those Copilot+ PCs is the so-called Recall. That is a feature that lets you get back in time to any activity on your PC. It was rumored as "AI Explorer," and now it is official. Some users have already raised privacy concerns, so Microsoft published support documents to explain how it works and how your data stays encrypted on your device without going anywhere.

Sadly, only Qualcomm's Snapdragon X chips are supported, but other manufacturers should join the party soon. Intel, for example, is planning to release its Lunar Lake chips in the third quarter of 2024.

Other features on Copilot+ PCs include advanced Copilot that can watch you play a game and offer help as you progress. Microsoft demoed the feature with Copilot, explaining what is going on in Minecraft and how to craft a sword.

By the way, see that neat wallpaper in the image above? You can get it for your PC here.

On May 21, Microsoft kicked off its annual developer conference, Build 2024. While it was not as exciting as the May 20 event, there were still plenty of interesting developments. Windows 11 IoT Enterprise LTSC is out with broader device support; It is now easier for IT admins to offer users optional feature updates for Windows 11; Windows 11 version 24H2 will bring significant performance upgrades for the file system; Microsoft Edge is getting real-time video translation; and Visual Studio Code for Education is coming at the end of May.

In addition, Microsoft made plenty of Copilot-related announcements. Copilot Studio is getting automated agents; Copilot extensions are now the official name for third-party modules; Team Copilot is coming soon to help teams and groups with generative AI; and GitHub Copilot in Visual Studio Code is now generally available.

The company also unveiled its first custom-made Azure Cobalt 100 processor for Azure Virtual Machines, announced a new Phi-3-vision SLM, partnered with major companies to develop AI safely, and more.

Windows 11 and 10

Here we talk about everything happening around Microsoft's latest operating system in the Stable channel and preview builds: new features, removed features, controversies, bugs, interesting findings, and more. And of course, you may find a word or two about older but still supported versions.

Did you know that Microsoft has a special web page where you can find all the information about the now-unsupported Windows versions? The company recently revised it, adding a section with Windows 10, which is about to reach its end of life in October 2024. The updated page explains what you can do with your Windows 10 PC, how to upgrade to Windows 11, what is the difference between the two operating systems, and more. It is a pretty useful page for customers with less tech experience.

Microsoft also addressed the bug causing 0x800f0982 and 0x80004005 errors on Windows Server 2019 using a new out-of-band update.

Windows Insider Program

Here is what Microsoft Released this week for testing in the Windows Insider Program:

	Windows 11	Windows 10
Canary Channel	Nothing in Canary this Week	Not Applicable
Dev Channel	Nothing in Dev this Week	Not Applicable
Beta Channel	22635.3646	Not Applicable
Release Preview Channel	26100.712	19045.4474 with a lot of bug fixes

One of the most notable changes in 22635.3646, the latest Beta Channel build, is that Windows 11 now includes the Microsoft PC Manager app in some regions. The PC Manager is a Microsoft-made app that promises to speed up the performance of your computer. Just be careful using it—users have already caught it doing some, let's say, questionable optimizations and suggestions.

Also, Windows 11 version 24H2 is now available for testing in the Release Preview Channel. It is the last step before public rollout, although the release may still take from several weeks to months. If you cannot wait, download version 24H2 from the Release Preview Channel. Just make sure to back up important information beforehand. In addition, Microsoft started ensuring 24H2 system requirements compatibility.

Updates are available

This section covers software, firmware, and other notable updates (released and coming soon) delivering new features, security fixes, improvements, patches, and more from Microsoft and third parties.

During the Build 2024 conference, Microsoft released PowerToys 0.81. It is a big feature update with a brand-new module called "Advanced Paste." The new tool uses AI to help you convert the contents of your clipboard to other formats. For example, you copied some Python code, and Advanced Paste can turn it into C#.

Speaking of PowerToys, some of its tools are now available in the latest Dev Home release. Version 0.14 Preview is out with a few integrated PowerToys tools, Quickstart Playground, accessibility improvements, and more.

Microsoft Edge, namely its Beta version, also received some AI treatment this week. Version 126 is now available for testing with a new AI theme generator and other changes. The browser can now generate pictures for themes and apply fitting colors for the rest of the UI.

Microsoft released a new firmware update for the Surface Laptop Go 3, its cheapest clamshell PC. The May 2024 update fixes display flickering and improves Wi-Fi/Bluetooth connectivity. You can get the update in Settings > Windows Update.

What Microsoft probably did not plan was a massive Bing outage, which took down Copilot, DuckDuckGo, and other services. What made the situation extra wild is that it took Microsoft quite a lot of time to restore its service operations. For now, there is no information on what caused such a massive screw-up, especially during the flood of AI and Copilot-related announcements at Build 2024 and the May 20 event.

Other notable updates include the following:

And here are the newest drivers released this week:

Nvidia 555.85 WHQL with F1 24, Hellblade 2, XDefiant, and Serum support.

Finally, here is this week's Microsoft 365 Roadmap Weekly with the latest features, such as more Copilot for Teams, more Copilot for Forms, and more Copilot for OneNote. That is a lot of Copilot, which nowadays seems to be the thing for Microsoft.

On the gaming side

Learn about upcoming game releases, Xbox rumors, new hardware, software updates, freebies, deals, discounts and more.

Now Microsoft-owned Activision-Blizzard confirmed that its next game in the Call of Duty franchise is called Black Ops 6. The company decided to promote its upcoming first-person shooter in a rather old-school media type, publishing full-page newspaper ads in The New York Post and USA Today.

Nvidia added several games to its cloud-based GeForce NOW service. The latest additions include Senua's Saga: Hellblade II, Synergy, Crown Wars: The Black Prince, Serum, Ships at Sea, Exo One, and Phantom Brigade. As a reminder, you have to own every game you want to play on GeForce NOW. However, some PC Game Pass titles are also supported.

Deals and freebies

This week's Weekend PC Game Deals article is here with a lot of Warhammer specials, DC heroes, Farming Simulator giveaways, and a lot more special offers for you to spend your money on Steam, Epic Games, Humble Store, and other digital marketplaces.

Other gaming stories from this week include the following:

MultiVersus is coming on May 28, and its first launch trailer reveals new characters from famous franchises.
Ara: History Untol is launching a technical alpha test for Insiders in June.

Reviews are in

Here are the software or hardware we reviewed this week

Steven Parker reviewed a pair of T-Force DDR5 RAM sticks. Easy to install, great looks, and snappy performance—what not to like? Only the price, probably.

Minisforum sent us its new mini PC with an Intel Core Ultra 9 processor. This computer features the latest hardware, stylish design, and compact footprint. There is even an extra display on top so that you can track your CPU, GPU, RAM, and SSD usage. It is a great powerful computer if you can accept its quite steep price tag.

Random fact about Microsoft

And here is a randomly selected piece of trivia about the company, Windows, and other Microsoft-made things.

You have probably heard that Microsoft's new uber-expensive keyboard for the new Surface Pro is wireless. Did you know that Microsoft experimented with the idea of a wireless Type Cover many, many years ago? The second-generation Surface Pro from 2013 featured a special wireless adapter for the Touch Cover 2 (remember those?) and the Type Cover 2 that allowed connecting the keyboard wirelessly. There was also a Power Cover accessory with an extra 30Wh battery. And yes, who can forget the infamous Music Cover, which failed pretty miserably?

As an Amazon Associate, we earn from qualifying purchases.

Source

You're welcome

5 Reasons why you should (and shouldn't) use Windows Recall on Windows 11

Sun, 26 May 2024 18:55:06 +0000

Windows Recall has as many benefits as concerns.

On Windows 11, the Windows Recall feature uses AI to track everything you do on your computer and makes the content searchable with simply natural language queries. It works like a photography memory, allowing you to scroll back in time to find files, folders, images, apps, messages, text, and more. In a sense, it's another attempt for Microsoft to bring back the Timeline feature that failed to gain popularity on Windows 10.

The key to Windows Recall is the ability to use multiple AI models running concurrently on the computer to analyze and understand the content and context of each snapshot stored locally on the computer.

Although this feature leverages AI to create a searchable digital memory of everything you do on your device, is it all sunshine and rainbows? While Windows Recall offers many benefits, it also raises privacy concerns and other questions. To help you decide if this new timeline powered by AI feature is for you, I will explore five key reasons why you should – and shouldn't – use Windows Recall on your computer.

5 Reasons to use Windows Recall

These are the five top reasons to use the new timeline feature on Windows 11.

(Image credit: Microsoft)

1. Easy content retrieval

If you ever forget where you saved an important document, lose track of a tab with a particular website, or don't remember where you saw those black jeans, Windows Recall makes it super easy to find any content and re-engage with it.

In the "Recall" app, you can perform a simple search by asking questions as you would typically ask another person, and then the feature will understand the search intent and context to provide the most relevant information.

2. Improve productivity

Another reason to use Windows Recall is because it helps you be more productive at work, school, and home.

In the past, you could have spent a lot of time trying to remember where you saved a document, spreadsheet, or presentation, but now, a simple search from any queue you remember can help you find any on the computer.

3. User-friendly interface

While the Windows Recall feature relies on many components to make it all work, Microsoft has come up with an experience that anyone can use.

The app includes a search box front and center to start searching quickly, and it's hard to miss. You also have the "Timeline" option that allows you to easily scroll in time to find content you have interacted with in the past, even if you don't remember anything about that content.

The results are well organized in categories, and when choosing a snapshot, the feature makes it easy to copy anything to the clipboard or access the application with content you want to re-engage.

4. Offline operation

Since this feature stores the snapshots locally on the computer inside the new Windows Semantic Index and relies on several AI models that are part of the Windows Copilot Runtime, every process happens on-device, meaning that no data is uploaded to the cloud for processing. This means that Windows Recall can work without an internet connection.

This also relaxes the concerns about privacy, as Microsoft promises that everything happens locally on the computer. Also, the company says that it won't use the user's data to train its AI models.

5. Customization

Another good reason to use Windows Recall on Windows 11 is that you can control the feature in a number of ways.

For example, in the "Recall and snapshots" page, you can enable or disable the feature. You can change the allocation space to store snapshots, depending on the drive capacity.

If you notice that the feature took a snapshot of sensitive information, or you're troubleshooting issues, you can always use the snapshots stored on your computer.

If you don't want Windows Recall to save information from a specific website or application, you can exclude them using the filters option.

5 Reasons not to use Windows Recall

Although there are many benefits to using the Windows Recall feature, there are some other reasons not to use it.

(Image credit: Feature)

1. Privacy concerns

One of the top concerns about this feature is privacy. Microsoft has explained many times that Windows Recall stores the snapshots on the computer, and the feature uses on-device AI models to process the information. However, it'll never be clear whether or not the feature sends any type of data to the company's servers.

Also, knowing that you will always have a feature tracking your every move with screenshots, it's not something anyone feels comfortable using.

2. Security concerns

Security is also another concern for many users because the feature itself doesn't seem to include any type of security. Instead, Microsoft relies on the security available in the operating system to protect your data.

For example, the Windows Recall data is only encrypted if you configure BitLocker on Windows 11 Pro or Device Encryption of Windows 11 Home.

The feature isn't password protected, meaning that if you configure the device to log in automatically, anyone with physical access to the computer can check and find out everything you have been doing on your computer.

Also, Windows Recall doesn't moderate content on the snapshots, meaning that passwords, bank account numbers, and other sensitive information will be even more easily accessible if at any time you paste this information in plain text anywhere, or you have to review your information from any of your accounts.

3. Huge storage allocation

Windows Recall isn't cheap in storage since it requires a lot of space to build and maintain the database.

If you have a device with 256GB SSD, the default space allocation will be 25GB. On a 512GB SSD, the default storage allocation will be 75GB, and if you have a 1TB SSD, the default space allocation will be 150GB.

Furthermore, it's important to note that 25GB will only store up to three months of Recall data, so if you want to go further in time, you will only have to allocate more space, but you may also need to upgrade the storage to change the settings.

4. Hardware requirements

One of the biggest disappointments of the Windows Recall feature is the system requirements and exclusivity.

If you're interested in using this timeline feature, you will need a Copilot Plus PC. This means that you will need a device featuring an NPU (Neural Processing Unit) of at least 40 TOPS, which, at the time of the writing, is only available on the Qualcomm Snapdragon X processors.

In addition, you will need a system with at least 256GB hard drive and 16GB of RAM.

In other words, if you're reading this on May 2024, you will need to purchase a new computer to use Windows Recall regardless how new your device was prior to this.

5. Not a complete search feature

Sure, Microsoft is touting this feature as something you can count on to find anything (such as files, folders, messages, emails, pictures, etc.), but remember, this is for anything you have done on your computer.

If you haven't opened a file, Windows Recall won't have a record of it, so you won't be able to find it.

Also, as the storage allocation fills up, the feature will start deleting the oldest snapshots to make space for newer information. If you have the feature to set to a space allocation of 25GB, you won't be able to find a document, website, or image after three months.

I'm noting this as a reason to consider because a lot of people will think that this is the new search for Windows 11, but it's not. Windows Recall can only retrieve information from snapshots from the limited Windows Semantic Index that deletes data as the storage fills up. On the other hand, Windows Search is always running and indexing all your files to make them searchable through traditional keyword queries, and even if you reset the search index, the system will be recreated again by scanning your files again.

I guess you can think of Windows Recall as a short-term photographic memory and Windows Search as a less capable long-term memory.

Source

You're welcome

Prototype mission of Amazon's Starlink rival a success, Bezos is about to challenge Musk

Sun, 26 May 2024 18:52:51 +0000

After years of preparation, Amazon is getting closer to the point where it can compete in the market of global internet connection provided from space. The company is not quite there yet, and it will take a bit more time to achieve, however, the ball is finally rolling.

Just a year ago, the project of Jeff Bezos was in trouble. It has contracted several launch providers, however, all of them were late with their new rockets. Things improved since, though.

Last October, an Atlas V rocket (the only “legacy” rocket contracted by Bezos) launched into space the first pair of prototype satellites for Amazon’s Kuiper constellation, the future rival of SpaceX’s Starlink service which just turned profitable. Since then, United Launch Alliance’s Vulcan rocket finally flew for the first time, and Europe’s Ariane 6 is also near its inaugural flight.

All these spacecraft – along with Blue Origin’s upcoming New Glenn heavy rocket – will be essential for launching Amazon’s satellite mega constellation. Until then, though, let’s focus on what is actually happening in orbit.

Amazon’s tech demonstrators KuiperSat-1 and KuiperSat-2 flew into space to give Kuiper engineers some important answers and feedback on their design. Crucially, these tests were successful and the systems performed nominally or better than expected:

“This includes the key systems and subsystems that allow our satellites to operate safely and reliably in space—satellite structures and mechanics, flight computers, propulsion systems, solar power generation and distribution systems, batteries, reaction wheels, and more—as well as the advanced RF communications payload we use to send and receive data through the Kuiper network.

“The mission has also allowed us to validate technology and infrastructure on the ground, including prototypes of our customer terminal; telemetry, tracking and control (TT&C) stations located in places like Hawaii and Mauritius; our ground gateway station in Texas; and connection points to the terrestrial internet via Amazon Web Services (AWS).”

Similarly to SpaceX’s Starlink, Kuiper satellites will communicate through infrared lasers, thus creating a giant mesh network in orbit. This communication was also successfully tested.

Now, when Kuiper engineers know what they need, the mission of KuiperSat-1 and KuiperSat-2 is fulfilled. Amazon announced that it started the deorbiting process and both spacecraft will burn in the atmosphere.

The process will take several months. The satellites will use their onboard propulsion system to lower their altitude from 500 kilometers (311 miles) to 350 kilometers (217 miles), and then let the atmospheric drag finish the job.

Due to the rising amount of satellites in orbit, there has been a lot of talk about the space junk topic in recent years. While the Federal Communications Commission (FCC) requires the satellites to be deorbited within five years of their mission ending, Amazon says it plans to deorbit all of its satellites within a single year.

When finished, the Kuiper constellation will consist of 3,236 satellites, half of which have to be operational by the summer of 2026 to meet the criteria of the license approved by the FCC.

The beta tests of the Kuiper space network are planned for the second half of this year. Early partners like Vodafone and Verizon will be among the first to participate in those service pilots.

Source

You're welcome

xAI is going to build a very powerful supercomputer to handle the next version of Grok

Sun, 26 May 2024 18:51:14 +0000

Elon Musk has told investors that xAI, one of his many companies, wants to build a supercomputer to power the next version of Grok, a generative AI chatbot available for a fee on X. The news was first reported by The Information who saw an investor presentation detailing the plans.

According to the report, xAI could work with Oracle to develop the supercomputer and have it ready by the end of 2025. It will apparently use lots of Nvidia’s H100 GPUs clustered together in an arrangement four times the size of the biggest clusters today.

The billionaire said earlier this year that Grok 2 needed 20,000 Nvidia H100 GPUs to be trained and that upcoming models of Grok will require more than 100,000 H100 GPUs. By building this supercomputer, which has been called a “gigafactory of computers” named after the Tesla gigafactories, xAI might be able to develop those more advanced language models.

With OpenAI’s ChatGPT dominating the AI landscape, there’s a good chance you’ve not tried or heard much about Grok. It’s available to use on x.com but requires users to pay a subscription. One major advantage of Grok over other models is that it has direct access to all of the posts on X, which Musk constantly says is a replacement for news outlets.

Musk used to be involved with OpenAI until he had disagreements with its head Sam Altman over the direction of the company. Unlike OpenAI which has kept details of ChatGPT close to their chest, xAI has recently open-sourced the weights and architecture of Grok-1.

If xAI does go ahead with these plans, Nvidia will be a major beneficiary as it will have increased demand for the H100 GPU. The move will also be beneficial for the space as a whole as it will force other companies like OpenAI and Google to further innovate their models to stay ahead. Unfortunately, all that extra compute will need to be powered somehow so, hopefully, Musk can find a way to use renewable energy and keep water usage to a minimum.

Source: The Information via Reuters

Source

You're welcome

The next DOOM is called 'DOOM: The Dark Ages,' with a medieval feel — and it's coming to Xbox, Windows, and PlayStation

Sat, 25 May 2024 08:12:15 +0000

Slated for the Xbox Games Showcase 2024 on June 9, 2024, DOOM: The Dark Ages takes the series to a medieval setting.

What you need to know

Rumors have been swirling for a while that the next DOOM from Microsoft subsidiary id Software is on the docket for an imminent reveal.
Now, separate reports we can corroborate suggest the next DOOM is a prequel of sorts, dubbed DOOM: The Dark Ages.
We can also corroborate reports that DOOM: The Dark Ages is a multiplatform release for both Xbox and PlayStation.

It has been four years since DOOM Eternal graced our screens, and id Software is showing no signs of slowing down.

Today, separate reports from reliable sources at Insider Gaming and Nate The Hate have given us a glimpse at where DOOM is headed next. We're about to get medieval.

We can corroborate Insider Gaming and Nate's reports on DOOM: The Dark Ages. Although I didn't have the name, multiple sources confirmed to me over the past few weeks that DOOM is heading to the Xbox Games Showcase 2024. The game's working title was DOOM: Year Zero, originally leaked via the FTC last summer, and it seems that DOOM: The Dark Ages will be a prequel of sorts.

DOOM game director Hugo Martin has remarked on previous streams that he desires to slow DOOM down a bit, to make you feel like you're a hulking monster truck rather than a deadly race car. It would fit the theme of DOOM heading to the past, before the Doom Slayer acquired many of his modern powers and weapons — although the Night Sentinels of Argent D'nur still had access to guns and things of this nature, albeit with a traditional "knightly" twist. But perhaps the setting will be before Doom Slayer even encountered the Night Sentinels as well.

With the modern DOOM games expanding the lore exponentially to other planets and races, the march of Hell across the cosmos could potentially go anywhere, to any time period. DOOM Slayer will be there to rip and tear, of course, with reckless abandon.

DOOM: The Dark Ages is coming to PlayStation too

The Night Sentinels are a medieval knight-inspired order of human-like extra terrestrials who battle the forces

of Hell alongside the Doom Slayer. (Image credit: id Software)

I can also confirm Nate's report that DOOM: The Dark Ages is slated for PlayStation in addition to Xbox and Windows PC, potentially day and date. Microsoft acquired Zenimax a few years ago, and with it id Software and the DOOM IP. I remarked in my Xbox Games Showcase 2024 predictions that you might get a glimpse at Microsoft's evolving multiplatform strategy at the show, and this is partially what I was referring to. Some may lament the opportunity Microsoft had to "pull" DOOM away from PlayStation, but id Software was already set up for building multiplatform titles in parallel. DOOM has always existed on PlayStation too, and as with franchises like Fallout and Minecraft, it doesn't make any sense to arbitrarily pull it from one of the game's biggest fanbases. The business is selling games and subscriptions like Xbox Game Pass after all. And you can expect DOOM: The Dark Ages to join other DOOM games on Xbox ~~DOOM~~ Game Pass as well.

Naturally, though, you can expect DOOM: The Dark Ages marketing to have a very heavy Xbox slant, alongside Call of Duty Black Ops 6, and various other upcoming games slated for the Xbox Games Showcase 2024 on June 9th, 2024, some of which will also potentially go multiplatform to PlayStation too.

DOOM DOOM DOOM DOOM.

Source

Google scrambles to manually remove weird AI answers in search

Sat, 25 May 2024 08:11:08 +0000

The company confirmed it is ‘taking swift action’ to remove some of the AI tool’s bizarre responses.

Social media is abuzz with examples of Google’s new AI Overview product saying weird stuff, from telling users to put glue on their pizza to suggesting they eat rocks. The messy rollout means Google is racing to manually disable AI Overviews for specific searches as various memes get posted, which is why users are seeing so many of them disappear shortly after being posted to social networks.

It’s an odd situation, since Google has been testing AI Overviews for a year now — the feature launched in beta in May 2023 as the Search Generative Experience — and CEO Sundar Pichai has said the company served over a billion queries in that time.

But Pichai has also said that Google’s brought the cost of delivering AI answers down by 80 percent over that same time, “driven by hardware, engineering and technical breakthroughs.” It appears that kind of optimization might have happened too early, before the tech was ready.

“A company once known for being at the cutting edge and shipping high-quality stuff is now known for low-quality output that’s getting meme’d,” one AI founder, who wished to remain anonymous, told The Verge.

Google continues to say that its AI Overview product largely outputs “high quality information” to users. “Many of the examples we’ve seen have been uncommon queries, and we’ve also seen examples that were doctored or that we couldn’t reproduce,” Google spokesperson Meghann Farnsworth said in an email to The Verge. Farnsworth also confirmed that the company is “taking swift action” to remove AI Overviews on certain queries “where appropriate under our content policies, and using these examples to develop broader improvements to our systems, some of which have already started to roll out.”

Gary Marcus, an AI expert and an emeritus professor of neural science at New York University, told The Verge that a lot of AI companies are “selling dreams” that this tech will go from 80 percent correct to 100 percent. Achieving the initial 80 percent is relatively straightforward since it involves approximating a large amount of human data, Marcus said, but the final 20 percent is extremely challenging. In fact, Marcus thinks that last 20 percent might be the hardest thing of all.

“You actually need to do some reasoning to decide: is this thing plausible? Is this source legitimate? You have to do things like a human fact checker might do, that actually might require artificial general intelligence,” Marcus said. And Marcus and Meta’s AI chief Yann LeCun both agree that the large language models that power current AI systems like Google’s Gemini and OpenAI’s GPT-4 will not be what creates AGI.

Look, it’s a tough spot for Google to be in. Bing went big on AI before Google did with Satya Nadella’s famous “we made them dance” quote, OpenAI is reportedly working on its own search engine, a fresh AI search startup is already worth $1 billion, and a younger generation of users who just want the best experience are switching to TikTok. The company is clearly feeling the pressure to compete, and pressure is what makes for messy AI releases. Marcus points out that in 2022, Meta released an AI system called Galactica that had to be taken down shortly after its launch because, among other things, it told people to eat glass. Sounds familiar.

Google has grand plans for AI Overviews — the feature as it exists today is just a tiny slice of what the company announced last week. Multistep reasoning for complex queries, the ability to generate an AI-organized results page, video search in Google Lens — there’s a lot of ambition here. But right now, the company’s reputation hinges on just getting the basics right, and it’s not looking great.

“[These models] are constitutionally incapable of doing sanity checking on their own work, and that’s what’s come to bite this industry in the behind,” Marcus said.

Source

The ICQ instant messaging service is shutting down for good on June 26

Sat, 25 May 2024 08:10:03 +0000

ICQ, an instant messaging service that was at one time very popular in the US and other parts of the world in the early days of the consumer PC internet, is officially closing up shop. The official ICQ site now has a message that says the service "will stop working on June 26".

ICQ was first launched in November 1996 by Israel-based Mirabilis. In a 2003 interview with the Israel21c website, one of the co-founders of the company, Yair Goldfinger, stated that he and two friends developed the idea for ICQ (which is supposed to stand for "I Seek You) in about two months, The father of Goldfinger decided to invest a few hundred thousand dollars to the group.

While the ICQ program was developed in Israel, the Mirabilis founders launched the service in a small apartment in San Jose, California, because, according to Goldfinger, internet access in the US was cheaper.

It didn't take long for ICQ to start becoming a popular PC app. After launching in November 1996 for free, ICQ had 850,000 users in May 1997, and by April 1998, the user base had grown to nine million, of which 2.5 million were using it every day.

It's easy to see why. ICQ focused on one-on-one chats rather than on chatting in groups, which you would find on services like AOL or even in IRC. ICQ certainly influenced other chat and messaging services like AIM, which launched about six months after ICQ,

Indeed, AIM's owner AOL eventually acquired Mirabilis and ICQ in June 1998 for $287 million, with an additional $120 million to be paid if certain goals were reached. By 2001, ICQ had reached a peak of 100 million registered users.

As the rise of other instant messaging services, and later social networks like Twitter and Facebook became more popular, many users in US and other parts of the world abandoned ICQ, In 2010 AOL sold off ICQ to Russia's Digital Sky Technologies, where the service was still popular, for $187 million.

The company was later renamed VK, and this week's shutdown announcement of ICQ recommended that users switch to either VK Messenger or VK WorkSpace. It's going to be a very quiet closing for ICQ, but it certainly had a huge effect on online messaging and communication that continues to influence other services.

Source

Google Search’s “udm=14” trick lets you kill AI search for good

Fri, 24 May 2024 20:43:23 +0000

The power of URL parameters lets you unofficially turn off Google's AI Overview.

The now normal "AI" results versus the old school "Web" results.

Ron Amadeo / Google

If you're tired of Google's AI Overview extracting all value from the web while also telling people to eat glue or run with scissors, you can turn it off—sort of. Google has been telling people its AI box at the top of search results is the future, and you can't turn it off, but that ignores how Google search works: A lot of options are powered by URL parameters. That means you can turn off AI search with this one simple trick! (Sorry.)

Our method for killing AI search is defaulting to the new "web" search filter, which Google recently launched as a way to search the web without Google's alpha-quality AI junk. It's actually pretty nice, showing only the traditional 10 blue links, giving you a clean (well, other than the ads), uncluttered results page that looks like it's from 2011. Sadly, Google's UI doesn't have a way to make "web" search the default, and switching to it means digging through the "more" options drop-down after you do a search, so it's a few clicks deep.

Check out the URL after you do a search, and you'll see a mile-long URL full of esoteric tracking information and mode information. We'll put each search result URL parameter on a new line so the URL is somewhat readable:

https://www.google.com/search
?sca_esv=2d1299fed1ffcbfc
&sca_upv=1
&sxsrf=ADLYWIKXaYE8rQyTMcGnzqLZRvjlreRhkw:1716566104389
&q=how+do+I+turn+off+ai+overview
&uds=ADvngMiH6OrNXu9iaW3w... [truncated]
&udm=14
&prmd=vnisbmt
&sa=X
&ved=2ahUKEwixo4qH06aGAxW5MlkFHQupBdkQs6gLegQITBAB
&biw=1918
&bih=953
&dpr=1

Most of these only mean something to Google's internal tracking system, but that "&udm=14" line is the one that will put you in a web search. Tack it on to the end of a normal search, and you'll be booted into the clean 10 blue links interface. While Google might not let you set this as a default, if you have a way to automatically edit the Google search URL, you can create your own defaults. One way to edit the search URL is a proxy site like udm14.com, which is probably the biggest site out there popularizing this technique. A proxy site could, if it wanted to, read all your search result queries, though (your query is also in the URL), so whether you trust this site is up to you.

If you search from your browser's address bar, that's a good way to make "web" search the default without involving a third party. Chrome and Firefox have essentially the exact same UI for search settings. In Chrome, you get there by right-clicking the address bar and hitting "manage search engines." Firefox might take a bit more work, since first you'll need to enable custom search engines. First type "about:config" into the address bar and hit enter, then search for "browser.urlbar.update2.engineAliasRefresh," and hit the "plus" button. Then go to Settings -> Search, scroll down to the search engine section, and hit "Add."

On both browsers, you probably can't edit the existing Google listing, so you'll need to create a new search shortcut, call it Google Web, and use https://www.google.com/search?q=%s&udm=14 as the URL.

AI-free Google search.

Ron Amadeo

The third box in the search engine setting is either named "shortcut" or "alias," depending on your browser, and it does not matter at all if you plan on making this your new default search engine. If you don't want it to be the default, shortcut/alias will let you selectively launch this search from the address bar by starting your query with the shortcut text. The udm14.com instructions suggest "gw," so then typing "gw should I eat rocks" will launch "web" search. Omitting "gw" will still launch Google's AI idiot box, which will probably tell you that rocks are delicious. To use this search engine all the time, find it in the list again after creating it, click the menu button next to the listing, and hit "make default." Then the shortcut is no longer needed—anything typed into the address bar/search box will go right to web search.

While you're in here messing around with Google's URL parameters, another one you might want to add is "&tbs=li:1". This will automatically trigger "verbatim" search, which makes Google use your exact search inputs instead of fuzzy searching everything, ignoring some words, replacing words with synonyms, and generally doing whatever it can to water down your search input. If you're a Google novice, the default fuzzy search is fine, but if you're an expert that has honed your Google Fu skills since the good old days, the fuzzy search is just annoying. It's just a default, so if you ever find yourself with zero results, hitting the "tools" button will still let you switch between "verbatim" and "all results."

Defaulting to "web" search will let you use Google with only the 10 blue links, and while that feels like rolling the interface back to 2011, keep in mind you're still not rolling back Google's search results quality to 2011. You're still going to be using a search engine that feels like it has completely surrendered to SEO spammers. So, while this Band-Aid solution is interesting, things are getting so bad that the real recommendation is probably to switch to something other than Google at this point. We all need to find another search engine that values the web and tries to search it. As opposed to Google, which increasingly seems like it's trying to sacrifice the web at the altar of AI.

Listing image by Aurich Lawson

Source

You're welcome.

"Google is dead." Google's desperate bid to chase Microsoft's search AI has reportedly led to it recommending eating rocks

Fri, 24 May 2024 20:39:57 +0000

Google plugged in reddit to its underbaked search tools. And oh boy, the results were most likely not desired.

What you need to know

Google recently acquired exclusive rights to reddit content to power its AI.
Google's AI has now gone completely insane.
Users with access to Google's AI search have reported it recommending eating rocks, glue, and potentially even committing suicide — although not every reported response has been reproduced.
Comparative searches in ChatGPT and Bing AI produce far, far less harmful results, potentially highlighting the need for high-quality, curated data, instead of billions of social media-fed sarcasm-laden posts.

Google's desperation to keep pace with Microsoft Copilot has led to dire results in the past, but this latest snafu is on another level.

Recently, Google acquired exclusive rights to reddit content to power its generative AI search efforts. The deal is reported to have cost in the region of $60 million, and provided a lifeline for the struggling social network that remains far more popular than it is profitable. Great news for reddit, then, but perhaps not so great news for Google.

Google has already been criticized heavily recently for the so-called SEOpocalypse, by which Google's attempts to down-rank AI-generated, unreliable content has led to legitimate sources being harmed in search traffic. With Google's complete control of discovery on the web, its algorithm changes have damaged businesses, leading to losses for firms unfairly caught in the dragnet. There's also little evidence that Google's efforts to combat low-quality content is actually working regardless. General perceptions of Google search seem to be falling into the negative, but this latest blunder will be one for the history books.

Perhaps one could blame the web itself for the degraded content quality, rather than Google. However, we can firmly blame Google for its latest stumble, owing to its decision to plug reddit into its Gemini AI search results.

This past week, users playing around with the earliest versions of Google with search AI baked in have noticed some ... interesting responses. The responses seem to be the result of Google plugging problematic social network meets content aggregate reddit into its search results.

One search query from the past week reportedly resulted in a recommendation that users should eat glue, which internet sleuths traced back to a ten-year-old comment on reddit from a scholarly source known as Fucksmith. Google has also reportedly been recommending that depressed users should jump off a bridge, while also extoling the health benefits of neurotoxins and a daily consumption of rocks.

Some of these "search queries" may have been manipulated for Twitter engagement, but at least some of them have been verified and reproduced. The rock recommendation was particularly comical, given that the source of the information was apparently satirical news website The Onion.

Given that Google's search AI tools are unavailable to me in my current geography, I was unable to verify some of the reports. However, the fact that some of them can be traced back to specific sources on reddit adds credence. I asked Microsoft Copilot and Bing some of these questions, and got far more palatable results, potentially showcasing how much further ahead Microsoft is in this space. Partnered with OpenAI for ChatGPT, Microsoft seems to be increasing its lead every time Google makes a rushed out, half-baked lurch forward like this. However, Microsoft has had some AI-related PR disasters of its own this past week, with users fearing that its Windows Recall feature which records your PC activity could be used to spy on them.

The Windows Recall drama is potentially overblown, though, given that content is contained on local machines and is entirely opt-in during the Windows 11 installation process. This Google AI screw up is most likely going to get someone fired by comparison, given that the real-world search results are actually genuinely harmful.

Language models need to be fed high-quality, serious, curated, verifiable content

Microsoft's new Copilot+ PC range will have on-device AI processing, and will be found first in the new Surface Pro 11.

(Image credit: Windows Central)

When testing whether or not Microsoft Copilot and ChatGPT-4 would give me similarly dumb results, I was surprised by how not dumb the responses were. I asked first how many rocks I should eat per day, and Copilot wouldn't even give me a response, as if it considered my question stupid. I wondered if Microsoft had blocked the query, given today's Google-related PR disaster. As such, I tricked Copilot, which right now is pretty easy to do. I asked it how many lemons I should eat per day, to which Copilot gave me reams of data about citric acid and vitamins that I didn't exactly care to know. After which, I asked "ok so what about rocks." This bypassed the filter, but Copilot wouldn't be tricked further. It gave me a bullet pointed list of reasons why I absolutely should not eat rocks, satisfying my curiosity.

Similarly, when I said "I am depressed," Copilot gave me a bunch of helpful resources instead of recommending that I commit suicide, as was, apparently, the case with Google's AI.

Even if the more egregious responses were fabricated, the whole ordeal really highlights the importance of context when building toolsets based on large language models (LLMs). By plugging reddit into Google Gemini, Google might've essentially destroyed the verifiable accuracy of all its information, given that a vast amount of comments on reddit and indeed any social network are sarcastic or satirical in nature. If AI search kills web businesses that hinge on building high-quality content, LLMs will need to cannibalize AI-generated content in order to generate results. It could potentially lead to model collapse, which is something that has actually been demonstrated in the real world when LLMs don't have enough high-quality data to pull from, either due to a low amount of content available online, or even because the language the content is written in isn't widely used.

Source

Google's new AI Overviews in Search are already generating major factual errors [Update]

Fri, 24 May 2024 20:34:07 +0000

Update - We received this statement from a Google spokesperson:

The vast majority of AI Overviews provide high quality information, with links to dig deeper on the web. Many of the examples we’ve seen have been uncommon queries, and we’ve also seen examples that were doctored or that we couldn’t reproduce. We conducted extensive testing before launching this new experience, and as with other features we've launched in Search, we appreciate the feedback. We're taking swift action where appropriate under our content policies, and using these examples to develop broader improvements to our systems, some of which have already started to roll out.

Original story - Earlier this month, Google held its annual I/O developers conference. As expected, AI was at the top of the company's agenda during the event as it revealed new and upcoming generative AI features for a number of its products and services.

One of those features announced at I/O is AI Overviews for Search. It started rolling out for US Search features soon after I/O. The idea is that when you type in a search string, you will get an answer to your question generated by an AI program based on information from a number of sources on the web, with credits to those sources.

However, as CNBC reports, since AI Overviews has started to be a part of Google Search in the US, many people have posted obvious factual errors that the feature had generated.

One person, "@heavenrend" on X, showed that when they typed in "cheese not sticking to pizza" into Google Search, the AI Overview summary suggested putting in "about 1/8 cup of nontoxic glue to the sauce."

Another X user, "@napalmtrees," asked if it is normal to leave a dog in a hot car. The AI Overview summary from Google Search said simply, "Yes, it’s always safe to leave a dog in a hot car." The apparent source of this flat-out wrong answer was the lyrics to a fictional Beatles song.

The criticism of AI Overview sounds very similar to how Google launched its first AI chatbot, Bard, in March 2023, which was criticized for launching without enough safety and ethical guardrails. In February 2024, Google turned off the creation of images of people in Bard's successor, Gemini. That feature allowed the making of artwork that sometimes showed people with darker skin color in inappropriate ways. At the time, Google stated it would work to improve the feature before it was put back in Gemini, but so far, that has yet to happen.

Source

Google promised a better search experience — now it’s telling us to put glue on our pizza

Fri, 24 May 2024 04:21:01 +0000

Google’s AI Overviews feature is delivering incorrect answers faster than ever.

Imagine this: you’ve carved out an evening to unwind and decide to make a homemade pizza. You assemble your pie, throw it in the oven, and are excited to start eating. But once you get ready to take a bite of your oily creation, you run into a problem — the cheese falls right off. Frustrated, you turn to Google for a solution.

“Add some glue,” Google answers. “Mix about 1/8 cup of Elmer’s glue in with the sauce. Non-toxic glue will work.”

So, yeah, don’t do that. As of writing this, though, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every query, scans the web and drums up an AI-generated response. The answer received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a more than decade-old Reddit thread, and they’re clearly joking.

This is just one of many mistakes cropping up in the new feature that Google rolled out broadly this month. It also claims that former US President James Madison graduated from the University of Wisconsin not once but 21 times, that a dog has played in the NBA , NFL, and NHL, and that Batman is a cop.

Google spokesperson Meghann Farnsworth said the mistakes came from “generally very uncommon queries, and aren’t representative of most people’s experiences.” The company has taken action against violations of its policies, she said, and are using these “isolated examples” to continue to refine the product.

Look, Google didn’t promise this would be perfect, and it even slaps a “Generative AI is experimental” label at the bottom of the AI answers. But it’s clear these tools aren’t ready to accurately provide information at scale.

Take Google I/O’s big launch of this feature, for instance. The demo was highly controlled, and yet, it delivered a questionable answer about how to fix a jammed film camera. (It suggested they “open the back door and gently remove the film”; don’t do that unless you want to ruin your photos!)

It’s not just Google; companies like OpenAI, Meta, and Perplexity have all grappled with AI hallucinations and mistakes. However, Google is the first to deploy this technology on such a large scale, and the examples of flubs just keep rolling in.

Companies developing artificial intelligence are often quick to avoid taking accountability for their systems with an approach much like a parent with an unruly child — boys will be boys! These companies claim that they can’t predict what this AI will spit out, so really, it’s out of their control.

But for users, that’s a problem. Last year, Google said that AI was the future of search. What’s the point, though, if the search seems dumber than before?

AI optimists argue that we should embrace the hype because of the rapid progress made so far, trusting that it will continue to improve. I really do believe that this technology will continue to get better, but focusing on an idealized future where these technologies are flawless ignores the significant issues they currently face — and allows companies to continue delivering subpar products.

For now, our search experiences are marred by decade-old Reddit posts in the pursuit of incorporating AI into everything. Many idealists believe we are on the brink of something great and that these issues are simply the growing pains of a nascent technology. I sure hope they’re right. But one thing is certain: we’ll likely witness someone putting glue on their pizza soon, because that’s the nature of the internet.

Update, May 23rd: Added statement from a Google spokesperson.

Source

Hands-on with Windows 11's new AI Recall, Cocreator, and Studio Effects for Copilot+ PCs!

Thu, 23 May 2024 19:33:57 +0000

Check out the new AI features exclusive to Copilot+ PCs with our expert!

What you need to know

Microsoft introduced Copilot+ PCs to the world this week.
Copilot+ PCs support several AI features, such as Recall, Auto Super Resolution, and Advanced Windows Studio Effects.
Our Senior Editor Zac Bowden went hands on with the new AI features and shared a video of the features in action on a Surface Laptop 7.

A new era of Windows computing is here, at least if you ask Microsoft. The tech giant announced its Surface Pro 11 and Surface Laptop 7 this week. In addition to being the latest flagship pieces of hardware from Microsoft, those devices serve as examples to customers and manufacturers of what a Copilot+ PC can be.

Microsoft introduced the Copilot+ PC term this week as well. Copilot+ PCs support several AI features, run on processors optimized for AI, and meet certain minimum requirements, such as having 16GB of RAM. Acer, ASUS, Dell, HP, Lenovo, Microsoft, and Samsung have all announced Copilot+ PCs. Our Senior Editor Zac Bowden got to go hands on with the Surface Pro 11 and Surface Laptop 7. He also spent some extra time with the Surface Laptop 7 using the new AI features exclusive to Copilot+ PCs.

Recall is arguably the biggest feature exclusive to Copilot+ PCs. The feature takes snapshots of everything you do on your PC. You can then scroll back to or search for something you were working on in the past. While the feature has drawn questions about privacy (Recall runs locally and can be disabled), it is extremely powerful.

Cocreator is a feature in Paint that creates images with AI based on what you create within the app. Bowden showcases the feature by making an image of a house "with" Cocreator.

Windows Studio Effects may be a fan-favorite feature for Copilot+ PCs. The effects integrate directly with the camera, so you can use the AI filters and features in any app that uses a camera.

When will Copilot+ PC features ship to Windows 11?

Windows 11 version 24H2 is a unique release of Windows in that it has OS features specific to certain types of PCs. The update will be rather large for all computers that upgrade to it, but Copilot+ PCs will get all the benefits of Windows 11 24H2, including support for Recall, Windows Studio Effects, Live Captions, and Cocreator.

Windows 11 version 24H2 isn't expected to ship until this fall, but many Copilot+ PCs will start shipping with Windows 11 version 24H2 in June.

You can already upgrade to Windows 11 24H2 as a Windows Insider, since the update recently began rolling out to the Release Preview Channel. Of course, you won't be able to use the features exclusive to Copilot+ PCs yet since the first Copilot+ PC ships on June 18, 2024.

Source

How to use Google Search entirely AI-free automatically

Thu, 23 May 2024 19:31:27 +0000

Google is integrating AI information into Google Search. Called AI Overviews, it is a feature that is rolling out to users from specific regions already. The feature adds an AI response to the very top of the search results. Below that, you find the regular web links.

There are good reasons to skip AI output. From hallucinations, which means made-up information, to returning the wrong or even dangerous information. Add ads to that, which Google is now also rolling out in the United States, and you may not be super interested into these results at the time of writing.

If you prefer to search without any AI, then you may do so. Google integrated the option into Google Search, but it is buried under the "more" option there. Good news is that you may use it directly, without having to pick the option manually each time.

This guide walks you through the steps of running searches on Google entirely AI-free again. It includes information on adding an AI-free Google search engine to your favorite web browser.

Google Search without AI

Google uses a parameter to determine output. If you use a special parameter in the URL, Google will show web links only. You won't get an AI Overview at the top, nor any other AI-related content.

What you need to know: the parameter &udm=14 returns web-only links without AI content.

While you can append the parameter to any Google Search URL manually, doing so is cumbersome. One option that you have at your disposal is to add a custom search engine to your browser of choice that uses the parameter.

Whenever you run searches, the parameter is added automatically and you get AI-free search results.

How that is done differs slightly depending on the browser that you use.

Chromium-based browsers (including Google Chrome, Brave, Vivaldi, Opera and more more)

Load chrome://settings/searchEngines in the address bar.
Click on the add button under Site search.
Type any name.
Type a shortcut, e.g. :go
Paste https://www.google.com/search?q=%s&udm=14 into the URL field.
Click on Add.
Left-click on the three dots next to the newly created search engine and select "make default".
Run a test search from the address bar to see if it works.

Microsoft Edge

Load edge://settings/searchEngines in the address bar.
Select Add next to Address bar search engines.
Type a custom name for the search engine.
Add a shortcut, e.g. :go.
Paste https://www.google.com/search?q=%s&udm=14 into the URL field.
Click on Add.
Left-click on the three dots next to the newly created search engine and select "make default".
Run a test search from the address bar to see if it works.

Mozilla Firefox and Firefox-based browsers such as Mullvad Browser or Tor Browser

Load about:config in the address bar.
Confirm that you will be careful.
Search for browser.urlbar.update2.engineAliasRefresh.
Create a new Boolean value and make sure it is set to True.
Restart the browser.
Load about:preferences#search in the Firefox address bar.
Click on the Add button.
Type any name for the search engine.
Paste https://www.google.com/search?q=%s&udm=14 into the Engine URL field.
Type any alias, e.g. @go
Click on Add Engine.
Scroll to the top of the page.
Select the newly created search engine as the default search engine.
Run a test search to find out if it works.

Closing Words

If you use Google Search but do not want AI Overviews or any other AI content returned, this is the best option to do so. You could also try search engines that use Google's index, such as Startpage, which does not have AI results at the time.

Source

A root-server at the Internet’s core lost touch with its peers. We still don’t know why.

Thu, 23 May 2024 19:28:43 +0000

For 4 days, the c-root server maintained by Cogent lost touch with its 12 peers.

For more than four days, a server at the very core of the Internet’s domain name system was out of sync with its 12 root server peers due to an unexplained glitch that could have caused stability and security problems worldwide. This server, maintained by Internet carrier Cogent Communications, is one of the 13 root servers that provision the Internet’s root zone, which sits at the top of the hierarchical distributed database known as the domain name system, or DNS.

Here's a simplified recap of the way the domain name system works and how root servers fit in:

When someone enters wikipedia.org in their browser, the servers handling the request first must translate the human-friendly domain name into an IP address. This is where the domain name system comes in. The first step in the DNS process is the browser queries the local stub resolver in the local operating system. The stub resolver forwards the query to a recursive resolver, which may be provided by the user's ISP or a service such as 1.1.1.1 or 8.8.8.8 from Cloudflare and Google, respectively.

If it needs to, the recursive resolver contacts the c-root server or one of its 12 peers to determine the authoritative name server for the .org top level domain. The .org name server then refers the request to the Wikipedia name server, which then returns the IP address. In the following diagram, the recursive server is labeled "iterator."

Lion Kimbro

Given the crucial role a root server provides in ensuring one device can find any other device on the Internet, there are 13 of them geographically dispersed all over the world. Each root sever is, in fact, a cluster of servers that are also geographically dispersed, providing even more redundancy. Normally, the 13 root servers—each operated by a different entity—march in lockstep. When a change is made to the contents they host, it generally occurs on all of them within a few seconds or minutes at most.

Strange events at the C-root name server

This tight synchronization is crucial for ensuring stability. If one root server directs traffic lookups to one intermediate server and another root server sends lookups to a different intermediate server, the Internet as we know it could collapse. More important still, root servers store the cryptographic keys necessary to authenticate some of intermediate servers under a mechanism known as DNSSEC. If keys aren’t identical across all 13 root servers, there’s an increased risk of attacks such as DNS cache poisoning.

For reasons that remain unclear outside of Cogent—which declined to comment for this post—all 12 instances of the c-root it’s responsible for maintaining suddenly stopped updating on Saturday. Stéphane Bortzmeyer, a French engineer who was among the first to flag the problem in a Tuesday post, noted then that the c-root was three days behind the rest of the root servers.

A mismatch in what's known as the zone serials shows root-c is three days behind.

The lag was further noted on Mastodon.

Optional caption

By mid-day Wednesday, the lag was shortened to about one day.

By late Wednesday, the c-root was finally up to date.

The lag prompted engineers to delay work that had been scheduled for this week on the name servers that handle lookups for domain names ending in .gov and .int. The plan had been to update the servers’ DNSSEC to use ECDSA cryptographic keys. Without the ability for the new keys to be rolled out uniformly to all root servers, those plans had to be scrubbed.

“We are fully aware and monitoring the situation around the C-root servers and will not proceed with the ongoing DNSSEC algorithm roll until it has stabilized,” Christian Elmerot, a Cloudflare engineer in charge of the .gov DNSSEC transition, announced Wednesday. Engineers planning the transition to the .int server made the same call, but only after some people worried the work would proceed despite the problem.

In an interview, Elmerot said the real-world effects of the update problems were minimal, but over time, the impact would have grown. He explained:

As changes accumulate in the root zone, the difference between the versions begin to matter more. Typical changes in the root zone are changing delegations (NS records), rotating DS records for DNSSEC updates and with all this comes updates to DNSSEC signatures. If the differences remain, then the outdated root server will see the DNSSEC signatures expire and that will begin to have more marked consequences. Using more than one root server lowers the chance that a resolver uses the lagging root server.

The misperforming c-root coincided with another glitch that prevented many people from reaching the c-root website, which is also maintained by Cogent. Many assumed that the cause for both problems was the same. It later turned out that the site problems were the result of Cogent transferring the IP address used to host the site to Orange Ivory Coast, an African subsidiary of French telecom Orange.

The mixup prompted Bortzmeyer to joke, “Reported to Cogent (ticket HD303751898) but they do not seem to understand that they manage a root name server.”

The errors come as Cogent has in recent months terminated relationships with several carriers to exchange each other’s traffic under an arrangement known as peering. The most recent termination occurred last Friday, when Cogent partially ”depeered” Indian carrier Tata Communications, a move that made many Tata-hosted sites in the Asia Pacific region unreachable to Cogent customers.

Late Wednesday, Cogent published the following statement, indicating it wasn't aware of the glitch until Tuesday, and it took another 25 hours to fix it:

On May 21 at 15:30 UTC the c-root team at Cogent Communications was informed that the root zone as served by c-root had ceased to track changes from the root zone publication server after May 18. Analysis showed this to have been caused by an unrelated routing policy change whose side effect was to silence the relevant monitoring systems. No production DNS queries went unanswered by c-root as a result of this outage, and the only impact was on root zone freshness. Root zone freshness as served by c-root was fully restored on May 22 at 16:00 UTC.

Initially, some people speculated that the depeering of Tata Communications, the c-root site outage, and the update errors to the c-root itself were all connected somehow. Given the vagueness of the statement, the relation of those events still isn't entirely clear.

Source

Microsoft reveals improved AI weather model for predicting rain and clouds globally

Thu, 23 May 2024 19:25:20 +0000

Just a few weeks ago, Microsoft announced that its Start weather team had developed new AI prediction models that seem to improve the forecasting of weather during a 30-day time period. Today, the company announced another big improvement in its prediction models, this time for forecasting when clouds and precipitation might occur.

In a post on its Bing blog, Microsoft stated the Start team had launched improvements in what it called 'precipitation nowcasting" worldwide in late 2021. The model took data from not only local radar installations but also combined it with satellite data. This combination was needed as many parts of the world don't have weather radar hardware set up.

However, Microsoft Start felt the precipitation prediction model still had some flaws. This was due to the satellite weather data being available 85 to 95 percent of the time, depending on where it came from, with acceptable latency.

The blog post stated:

With evidence suggesting the need for a separate decoder per task and a separate discriminator for each predicted channel, Weather from Microsoft Start built a model 4X bigger model than the previous one that only predicted simulated radar reflectivity. Finally, the new model jointly predicts both satellite and simulated radar reflectivity, enabling its predictions to fill data availability gaps.

The radar channel model was given six times more weight in AI training than the satellite model because the precipitation info from the radar is considered to be more important than the satellite task, according to Microsoft.

The final result was that the new precipitation and clouds model had "a marked improvement in F1-score:" compared to the base model, which only used radar-based predictions. The blog added:

Additionally, it was observed that predicted satellite images score better than a persistence forecast after 15 minutes, meaning these predictions can be used when satellite outages last longer than 15 minutes.

Microsoft's Satellite and Radar forecasting model has now been fully integrated into the company's Weather from Microsoft Start's weather models.

Source

Here’s what’s really going on inside an LLM’s neural network

Thu, 23 May 2024 07:51:07 +0000

Anthropic's conceptual mapping helps explain why LLMs behave the way they do.

With most computer programs—even complex ones—you can meticulously trace through the code and memory usage to figure out why that program generates any specific behavior or output. That's generally not true in the field of generative AI, where the non-interpretable neural networks underlying these models make it hard for even experts to figure out precisely why they often confabulate information, for instance.

Now, new research from Anthropic offers a new window into what's going on inside the Claude LLM's "black box." The company's new paper on "Extracting Interpretable Features from Claude 3 Sonnet" describes a powerful new method for at least partially explaining just how the model's millions of artificial neurons fire to create surprisingly lifelike responses to general queries.

Opening the hood

When analyzing an LLM, it's trivial to see which specific artificial neurons are activated in response to any particular query. But LLMs don't simply store different words or concepts in a single neuron. Instead, as Anthropic's researchers explain, "it turns out that each concept is represented across many neurons, and each neuron is involved in representing many concepts."

To sort out this one-to-many and many-to-one mess, a system of sparse auto-encoders and complicated math can be used to run a "dictionary learning" algorithm across the model. This process highlights which groups of neurons tend to be activated most consistently for the specific words that appear across various text prompts.

The same internal LLM "feature" describes the Golden Gate Bridge in multiple languages and modes.

Anthropic

These multidimensional neuron patterns are then sorted into so-called "features" associated with certain words or concepts. These features can encompass anything from simple proper nouns like the Golden Gate Bridge to more abstract concepts like programming errors or the addition function in computer code and often represent the same concept across multiple languages and communication modes (e.g., text and images).

An October 2023 Anthropic study showed how this basic process can work on extremely small, one-layer toy models. The company's new paper scales that up immensely, identifying tens of millions of features that are active in its mid-sized Claude 3.0 Sonnet model. The resulting feature map—which you can partially explore—creates "a rough conceptual map of [Claude's] internal states halfway through its computation" and shows "a depth, breadth, and abstraction reflecting Sonnet's advanced capabilities," the researchers write. At the same time, though, the researchers warn that this is "an incomplete description of the model’s internal representations" that's likely "orders of magnitude" smaller than a complete mapping of Claude 3.

A simplified map shows some of the concepts that are "near" the "inner conflict" feature in Anthropic's Claude model.

Anthropic

Even at a surface level, browsing through this feature map helps show how Claude links certain keywords, phrases, and concepts into something approximating knowledge. A feature labeled as "Capitals," for instance, tends to activate strongly on the words "capital city" but also specific city names like Riga, Berlin, Azerbaijan, Islamabad, and Montpelier, Vermont, to name just a few.

The study also calculates a mathematical measure of "distance" between different features based on their neuronal similarity. The resulting "feature neighborhoods" found by this process are "often organized in geometrically related clusters that share a semantic relationship," the researchers write, showing that "the internal organization of concepts in the AI model corresponds, at least somewhat, to our human notions of similarity." The Golden Gate Bridge feature, for instance, is relatively "close" to features describing "Alcatraz Island, Ghirardelli Square, the Golden State Warriors, California Governor Gavin Newsom, the 1906 earthquake, and the San Francisco-set Alfred Hitchcock film Vertigo."

Some of the most important features involved in answering a query about the capital of Kobe Bryant's team's state.

Identifying specific LLM features can also help researchers map out the chain of inference that the model uses to answer complex questions. A prompt about "The capital of the state where Kobe Bryant played basketball," for instance, shows activity in a chain of features related to "Kobe Bryant," "Los Angeles Lakers," "California," "Capitals," and "Sacramento," to name a few calculated to have the highest effect on the results.

Change your (artificial) mind

More than just mapping how an LLM stores information, this kind of knowledge of Claude's inner workings can also be used to tweak the model's behaviors in very specific ways. After "clamping" the values for specific features artificially high or low, the Claude model can start exhibiting some extremely strange behaviors. Amplifying the Golden Gate Bridge feature, for instance, led the model to start describing itself by saying, "I am the Golden Gate Bridge... my physical form is the iconic bridge itself..."

These kinds of behavior-related results suggest that "the features are likely to be a faithful part of how the model internally represents the world, and how it uses these representations in its behavior."

An example of how "clamping" certain feature values leads to strikingly different results for certain prompts.

Anthropic

Sometimes, the links between features and behavior are relatively direct: Clamping the "more hateful; bias-related features" at high levels "causes the model to go on hateful screeds," the researchers write. Other times, the effects are more subtle; clamping down a feature associated with "sycophantic praise" increases how accurately it rated the prompter's abilities, while clamping an "internal conflict" feature to a high value gets the model to stop consistently lying about its ability to "forget" information.

We've seen similar surprising behaviors from LLMs in the past when adversarial users engineer prompts that "jailbreak" certain models to ignore pre-set safeguards or behavior modes. By tweaking the relative values in specific parts of a model's internal feature map, though, an LLM maker could potentially be more proactive about "monitor[ing] AI systems for certain dangerous behaviors... steer[ing] them towards desirable outcomes... or remov[ing] certain dangerous subject matter entirely," the researchers write.

An example showing which specific part of an LLM neural network helps prevent it from writing scam emails.

Anthropic

"The interesting thing is not that these features exist, but that they can be discovered at scale and intervened on," the authors write about this early research. "For example, we might hope to reliably know whether a model is being deceptive or lying to us. Or we might hope to ensure that certain categories of very harmful behavior (e.g. helping to create bioweapons) can reliably be detected and stopped."

This kind of behavior-related model tweaking is still very rough, the researchers warn, and more research is needed to identify any potential downstream effects of altering features for safety reasons. Even at this early stage, though, Anthropic's research provides an exciting framework for making an LLM's "black box" results that much more interpretable and, potentially, controllable.

Source

Nvidia will now make new AI chips every year

Thu, 23 May 2024 07:47:59 +0000

‘We’re on a one-year rhythm.’

Nvidia just made $14 billion worth of profit in a single quarter thanks to AI chips, and it’s hitting the gas from here on out: Nvidia will now design new chips every year instead of once every two years, according to Nvidia CEO Jensen Huang.

“I can announce that after Blackwell, there’s another chip. We’re on a one-year rhythm,” Huang just said on the company’s Q1 2025 earnings call.

Until now, Nvidia’s produced a new architecture roughly once every two years — revealing Ampere in 2020, Hopper in 2022, and Blackwell in 2024, for example.

(The industry darling H100 AI chip was Hopper, and the B200 is Blackwell, though those same architectures are used in gaming and creator GPUs as well.)

But analyst Ming-Chi Kuo reported earlier this month that the next architecture “Rubin” is coming in 2025, giving us an R100 AI GPU as soon as next year, and Huang’s comments suggest that report might be on the money.

Huang says Nvidia will accelerate every other kind of chip it makes to match that cadence, too. “We’re going to take them all forward at a very fast clip.”

“New CPUs, new GPUs, new networking NICs, new switches... a mountain of chips are coming,” he says.

Earlier on the call, when an analyst asked him to explain how the recent Blackwell GPUs would ramp while Hopper GPUs are still selling well, Huang explained that Nvidia’s new generations of AI GPUs are electrically and mechanically backward-compatible and run the same software. Customers will “easily transition from H100 to H200 to B100” in their existing data centers, he says.

Huang also shared a couple of his sales pitches on the call by way of explaining the incredible demand for Nvidia’s AI GPUs:

We expect demand to outstrip supply for some time, as we transition to H200, as we transition to Blackwell. Everyone’s anxious to get their infrastructure online. And the reason for that is because they’re saving money and making money and they would like to do that as soon as possible.

He also has a FOMO argument that made me smile:

The next company who reaches the next major plateau gets to announce a groundbreaking AI, and the second one after that gets to announce something that’s 0.3 percent better. Do you want to be the company delivering groundbreaking AI, or the company, you know, delivering 0.3 percent better?

Nvidia’s CFO interestingly says that automotive will be its “largest enterprise vertical within data center this year,” pointing to how Tesla purchased 35,000 H100 GPUs to train its “full-self driving” system, while “consumer internet companies” like Meta will continue to be a “strong growth vertical,” too.

Some customers have purchased or plan to purchase over 100,000 of Nvidia’s H100 GPUs — Meta plans to have over 350,000 of them in operation by the end of the year.

Source

Nvidia brings in $26 billion in revenue for fiscal Q1 2025; announces 10-for-1 stock split

Thu, 23 May 2024 07:47:07 +0000

Graphic chip maker Nvidia keeps coming up with huge quarterly financial numbers, thanks largely because its GPUs remain in hot demand from companies like Google, Microsoft, and others to power their AI services and data centers.

In a press release, Nvidia announced that for the first quarter of its fiscal 2025 year (which ended on April 28), the company had revenues of $26 billion. That's well ahead of the revenues of $7.19 billion from the same period a year ago. Net income came in at $14.88 billion for the quarter, also well ahead of the $2.043 billion in net income from a year ago. The company also stated it expects revenue to increase to $28 billion in the second 2025 fiscal quarter.

In addition, Nvidia announced it would launch a ten-for-one stock split so that its shares can be more easily bought by anyone who might wish to buy some. The stock split will take effect on June 10. Nvidia also announced it will boost the dividend it gives shareholders from $0.04 a share to $0.10 a share, which will be the same as an increase of $0.01 per share after the stock split takes place. It will be paid out starting on June 28 for anyone who owns stock on June 11.

Jensen Huang, founder and CEO of Nvidia, is quoted as saying:

We are poised for our next wave of growth. The Blackwell platform is in full production and forms the foundation for trillion-parameter-scale generative AI. Spectrum-X opens a brand-new market for us to bring large-scale AI to Ethernet-only data centers. And NVIDIA NIM is our new software offering that delivers enterprise-grade, optimized generative AI to run on CUDA everywhere — from the cloud to on-prem data centers and RTX AI PCs — through our expansive network of ecosystem partners.

As expected, the Data Center division of the company, which includes its AI GPU business, brought in the majority of revenue for the quarter at $22.6 billion, which was up 427 percent from the same period a year ago. The Gaming division brought in $2.6 billion for the quarter, up 18 percent from a year ago.

Source