microsoft

critics-question-tech-heavy-lineup-of-new-homeland-security-ai-safety-board

Critics question tech-heavy lineup of new Homeland Security AI safety board

AI, AI ethics, AI safety, alphabet, Amazon, Anthropic, Biz & IT, DHS, Dr. Margaret Mitchell, executive order, Google, Joe Biden, machine learning, Meta, microsoft, NVIDIA, openai, Policy, President Biden, sam altman, Satya Nadella, Sundar Pichai / /

Adventures in 21st century regulation —

CEO-heavy board to tackle elusive AI safety concept and apply it to US infrastructure.

–

A modified photo of a 1956 scientist carefully bottling

On Friday, the US Department of Homeland Security announced the formation of an Artificial Intelligence Safety and Security Board that consists of 22 members pulled from the tech industry, government, academia, and civil rights organizations. But given the nebulous nature of the term “AI,” which can apply to a broad spectrum of computer technology, it’s unclear if this group will even be able to agree on what exactly they are safeguarding us from.

President Biden directed DHS Secretary Alejandro Mayorkas to establish the board, which will meet for the first time in early May and subsequently on a quarterly basis.

The fundamental assumption posed by the board’s existence, and reflected in Biden’s AI executive order from October, is that AI is an inherently risky technology and that American citizens and businesses need to be protected from its misuse. Along those lines, the goal of the group is to help guard against foreign adversaries using AI to disrupt US infrastructure; develop recommendations to ensure the safe adoption of AI tech into transportation, energy, and Internet services; foster cross-sector collaboration between government and businesses; and create a forum where AI leaders to share information on AI security risks with the DHS.

It’s worth noting that the ill-defined nature of the term “Artificial Intelligence” does the new board no favors regarding scope and focus. AI can mean many different things: It can power a chatbot, fly an airplane, control the ghosts in Pac-Man, regulate the temperature of a nuclear reactor, or play a great game of chess. It can be all those things and more, and since many of those applications of AI work very differently, there’s no guarantee any two people on the board will be thinking about the same type of AI.

This confusion is reflected in the quotes provided by the DHS press release from new board members, some of whom are already talking about different types of AI. While OpenAI, Microsoft, and Anthropic are monetizing generative AI systems like ChatGPT based on large language models (LLMs), Ed Bastian, the CEO of Delta Air Lines, refers to entirely different classes of machine learning when he says, “By driving innovative tools like crew resourcing and turbulence prediction, AI is already making significant contributions to the reliability of our nation’s air travel system.”

So, defining the scope of what AI exactly means—and which applications of AI are new or dangerous—might be one of the key challenges for the new board.

A roundtable of Big Tech CEOs attracts criticism

For the inaugural meeting of the AI Safety and Security Board, the DHS selected a tech industry-heavy group, populated with CEOs of four major AI vendors (Sam Altman of OpenAI, Satya Nadella of Microsoft, Sundar Pichai of Alphabet, and Dario Amodei of Anthopic), CEO Jensen Huang of top AI chipmaker Nvidia, and representatives from other major tech companies like IBM, Adobe, Amazon, Cisco, and AMD. There are also reps from big aerospace and aviation: Northrop Grumman and Delta Air Lines.

Upon reading the announcement, some critics took issue with the board composition. On LinkedIn, founder of The Distributed AI Research Institute (DAIR) Timnit Gebru especially criticized OpenAI’s presence on the board and wrote, “I’ve now seen the full list and it is hilarious. Foxes guarding the hen house is an understatement.”

Critics question tech-heavy lineup of new Homeland Security AI safety board Read More »

putting-microsoft’s-cratering-xbox-console-sales-in-context

Putting Microsoft’s cratering Xbox console sales in context

gaming, microsoft, Nintendo, PlayStation, ps4, ps5, Sony, Switch, xbox / /

Down but not out —

Why declining quarterly numbers might not be awful news for Microsoft’s gaming business.

–

Scale is important, especially when talking about relative console sales.

Enlarge / Scale is important, especially when talking about relative console sales.

Aurich Lawson | Getty Images

Yesterday, Microsoft announced that it made 31 percent less off Xbox hardware in the first quarter of 2024 (ending in March) than it had the year before, a decrease it says was “driven by lower volume of consoles sold.” And that’s not because the console sold particularly well a year ago, either; Xbox hardware revenue for the first calendar quarter of 2023 was already down 30 percent from the previous year.

Those two data points speak to a console that is struggling to substantially increase its player base during a period that should, historically, be its strongest sales period. But getting wider context on those numbers is a bit difficult because of how Microsoft reports its Xbox sales numbers (i.e., only in terms of quarterly changes in total console hardware revenue). Comparing those annual shifts to the unit sales numbers that Nintendo and Sony report every quarter is not exactly simple.

Context clues

Significant declines in Xbox hardware revenue for four of the last five quarters stand out relative to competitors' unit sales.

Enlarge / Significant declines in Xbox hardware revenue for four of the last five quarters stand out relative to competitors’ unit sales.

Kyle Orland

To attempt some direct contextual comparison, we took unit sales numbers for some recent successful Sony and Nintendo consoles and converted them to Microsoft-style year-over-year percentage changes (aligned with the launch date for each console). For this analysis, we skipped over each console’s launch quarter, which contains less than three months of total sales (and often includes a lot of pent-up early adopter demand). We also skipped the first four quarters of a console’s life cycle, which don’t have a year-over-year comparison point from 12 months prior.

This still isn’t a perfect comparison. Unit sales don’t map directly to total hardware revenue due to things like inflation, remainder sales of Xbox One hardware, and price cuts/discounts (though the Xbox Series S/X, PS5, and Switch still have yet to see official price drops). It also doesn’t take into account the baseline sales levels from each console’s first year of sales, making total lifetime sales performance on the Xbox side hard to gauge (though recent data from a Take-Two investment call suggests the Xbox Series S/X has been heavily outsold by the PS5, at this point).

Even with all those caveats, the comparative data trends are pretty clear. At the start of their fourth full year on the market, recent successful consoles have been enjoying a general upswing in their year-over-year sales. Microsoft stands out as a major outlier, making less revenue from Xbox hardware in four of the last five quarters on a year-over-year basis.

Falling like dominoes.

Enlarge / Falling like dominoes.

Aurich Lawson

Those numbers suggest that the hardware sales rate for the Xbox Series S/X may have already peaked in the last year or two. That would be historically early for a console of this type; previous Ars analyses have shown PlayStation consoles generally see their sales peaks in their fourth or fifth year of life, and Nintendo portables have shown a similar sales trend, historically. The Xbox Series S/X progression, on the other hand, looks more similar to that of the Wii U, which was already deep in a “death spiral” at a similar point in its commercial life.

This is not the end

In the past, console sales trends like these would have been the sign of a hardware maker’s wider struggles to stay afloat in the gaming business. However, in today’s gaming market, Microsoft is in a place where console sales are not strictly required for overall success.

For instance, Microsoft’s total gaming revenue for the latest reported quarter was up 51 percent, thanks in large part to the “net impact from the Activision Blizzard acquisition.” Even before that (very expensive) merger was completed, Microsoft’s total gaming revenue was often partially buoyed by “growth in Game Pass” and strong “software content” sales across PC and other platforms.

Owning Call of Duty means being one of the biggest PS5 game publishers almost by definition.

Enlarge / Owning Call of Duty means being one of the biggest PS5 game publishers almost by definition.

Activision

Perhaps it’s no surprise that Microsoft has shown increasing willingness to take some former Xbox console exclusives to other platforms in recent months. In fact, following the Activision/Blizzard merger, Microsoft is now publishing more top-sellers on the PS5 than Sony. And let’s not forget the PC market, where Microsoft continues to sell millions of games above and beyond its PC Game Pass subscription business.

So, while the commercial future of Xbox hardware may look a bit uncertain, the future of Microsoft’s overall gaming business is in much less dire straits. That would be true even if Microsoft’s Xbox hardware revenue fell by 100 percent.

Putting Microsoft’s cratering Xbox console sales in context Read More »

tech-brands-are-forcing-ai-into-your-gadgets—whether-you-asked-for-it-or-not

Tech brands are forcing AI into your gadgets—whether you asked for it or not

AI, chatgpt, generative ai, Logitech, microsoft, Mouse, Nothing, Tech / /
Tech brands love hollering about the purported thrills of AI these days.

Enlarge / Tech brands love hollering about the purported thrills of AI these days.

Logitech announced a new mouse last week. A company rep reached out to inform Ars of Logitech’s “newest wireless mouse.” The gadget’s product page reads the same as of this writing.

I’ve had good experience with Logitech mice, especially wireless ones, one of which I’m using now. So I was keen to learn what Logitech might have done to improve on its previous wireless mouse designs. A quieter click? A new shape to better accommodate my overworked right hand? Multiple onboard profiles in a business-ready design?

I was disappointed to learn that the most distinct feature of the Logitech Signature AI Edition M750 is a button located south of the scroll wheel. This button is preprogrammed to launch the ChatGPT prompt builder, which Logitech recently added to its peripherals configuration app Options+.

That’s pretty much it.

Beyond that, the M750 looks just like the Logitech Signature M650, which came out in January 2022.  Also, the new mouse’s forward button (on the left side of the mouse) is preprogrammed to launch Windows or macOS dictation, and the back button opens ChatGPT within Options+. As of this writing, the new mouse’s MSRP is $10 higher ($50) than the M650’s.

  • The new M750 (pictured) is 4.26×2.4×1.52 inches and 3.57 ounces.

    Logitech

  • The M650 (pictured) comes in 3 sizes. The medium size is 4.26×2.4×1.52 inches and 3.58 ounces.

    Logitech

I asked Logitech about the M750 appearing to be the M650 but with an extra button, and a spokesperson responded by saying:

M750 is indeed not the same mouse as M650. It has an extra button that has been preprogrammed to trigger the Logi AI Prompt Builder once the user installs Logi Options+ app. Without Options+, the button does DPI toggle between 1,000 and 1,600 DPI.

However, a reprogrammable button south of a mouse’s scroll wheel that can be set to launch an app or toggle DPI out of the box is pretty common, including among Logitech mice. Logitech’s rep further claimed to me that the two mice use different electronic components, which Logitech refers to as the mouse’s platform. Logitech can reuse platforms for different models, the spokesperson said.

Logitech’s rep declined to comment on why the M650 didn’t have a button south of its scroll wheel. Price is a potential reason, but Logitech also sells cheaper mice with this feature.

Still, the minimal differences between the two suggest that the M750 isn’t worth a whole product release. I suspect that if it weren’t for Logitech’s trendy new software feature, the M750 wouldn’t have been promoted as a new product.

The M750 also raises the question of how many computer input devices need to be equipped with some sort of buzzy, generative AI-related feature.

Logitech’s ChatGPT prompt builder

Logitech’s much bigger release last week wasn’t a peripheral but an addition to its Options+ app. You don’t need the “new” M750 mouse to use Logitech’s AI Prompt Builder; I was able to program my MX Master 3S to launch it. Several Logitech mice and keyboards support AI Prompt Builder.

When you press a button that launches the prompt builder, an Options+ window appears. There, you can input text that Options+ will use to create a ChatGPT-appropriate prompt based on your needs:

A Logitech-provided image depicting its AI Prompt Builder software feature.

Enlarge / A Logitech-provided image depicting its AI Prompt Builder software feature.

Logitech

After you make your choices, another window opens with ChatGPT’s response. Logitech said the prompt builder requires a ChatGPT account, but I was able to use GPT-3.5 without entering one (the feature can also work with GPT-4).

The typical Arsian probably doesn’t need help creating a ChatGPT prompt, and Logitech’s new capability doesn’t work with any other chatbots. The prompt builder could be interesting to less technically savvy people interested in some handholding for early ChatGPT experiences. However, I doubt if people with an elementary understanding of generative AI need instant access to ChatGPT.

The point, though, is instant access to ChatGPT capabilities, something that Logitech is arguing is worthwhile for its professional users. Some Logitech customers, though, seem to disagree, especially with the AI Prompt Builder, meaning that Options+ has even more resources in the background.

But Logitech isn’t the only gadget company eager to tie one-touch AI access to a hardware button.

Pinching your earbuds to talk to ChatGPT

Similarly to Logitech, Nothing is trying to give its customers access to ChatGPT quickly. In this case, access occurs by pinching the device. This month, Nothing announced that it “integrated Nothing earbuds and Nothing OS with ChatGPT to offer users instant access to knowledge directly from the devices they use most, earbuds and smartphones.” The feature requires the latest Nothing OS and for the users to have a Nothing phone with ChatGPT installed. ChatGPT gestures work with Nothing’s Phone (2) and Nothing Ear and Nothing Ear (a), but Nothing plans to expand to additional phones via software updates.

Nothing's Ear and Ear (a) earbuds.

Enlarge / Nothing’s Ear and Ear (a) earbuds.

Nothing

Nothing also said it would embed “system-level entry points” to ChatGPT, like screenshot sharing and “Nothing-styled widgets,” to Nothing smartphone OSes.

A peek at setting up ChatGPT integration on the Nothing X app.

Enlarge / A peek at setting up ChatGPT integration on the Nothing X app.

Nothing’s ChatGPT integration may be a bit less intrusive than Logitech’s since users who don’t have ChatGPT on their phones won’t be affected. But, again, you may wonder how many people asked for this feature and how reliably it will function.

Tech brands are forcing AI into your gadgets—whether you asked for it or not Read More »

apple-releases-eight-small-ai-language-models-aimed-at-on-device-use

Apple releases eight small AI language models aimed at on-device use

AI, Apple, Biz & IT, chatgpt, chatgtp, CoreNet, IPhone, Llama 3, machine learning, Meta, microsoft, openai, OpenELM, Phi-3, small language models / /

Inside the Apple core —

OpenELM mirrors efforts by Microsoft to make useful small AI language models that run locally.

–

An illustration of a robot hand tossing an apple to a human hand.

Getty Images

In the world of AI, what might be called “small language models” have been growing in popularity recently because they can be run on a local device instead of requiring data center-grade computers in the cloud. On Wednesday, Apple introduced a set of tiny source-available AI language models called OpenELM that are small enough to run directly on a smartphone. They’re mostly proof-of-concept research models for now, but they could form the basis of future on-device AI offerings from Apple.

Apple’s new AI models, collectively named OpenELM for “Open-source Efficient Language Models,” are currently available on the Hugging Face under an Apple Sample Code License. Since there are some restrictions in the license, it may not fit the commonly accepted definition of “open source,” but the source code for OpenELM is available.

On Tuesday, we covered Microsoft’s Phi-3 models, which aim to achieve something similar: a useful level of language understanding and processing performance in small AI models that can run locally. Phi-3-mini features 3.8 billion parameters, but some of Apple’s OpenELM models are much smaller, ranging from 270 million to 3 billion parameters in eight distinct models.

In comparison, the largest model yet released in Meta’s Llama 3 family includes 70 billion parameters (with a 400 billion version on the way), and OpenAI’s GPT-3 from 2020 shipped with 175 billion parameters. Parameter count serves as a rough measure of AI model capability and complexity, but recent research has focused on making smaller AI language models as capable as larger ones were a few years ago.

The eight OpenELM models come in two flavors: four as “pretrained” (basically a raw, next-token version of the model) and four as instruction-tuned (fine-tuned for instruction following, which is more ideal for developing AI assistants and chatbots):

OpenELM features a 2048-token maximum context window. The models were trained on the publicly available datasets RefinedWeb, a version of PILE with duplications removed, a subset of RedPajama, and a subset of Dolma v1.6, which Apple says totals around 1.8 trillion tokens of data. Tokens are fragmented representations of data used by AI language models for processing.

Apple says its approach with OpenELM includes a “layer-wise scaling strategy” that reportedly allocates parameters more efficiently across each layer, saving not only computational resources but also improving the model’s performance while being trained on fewer tokens. According to Apple’s released white paper, this strategy has enabled OpenELM to achieve a 2.36 percent improvement in accuracy over Allen AI’s OLMo 1B (another small language model) while requiring half as many pre-training tokens.

An table comparing OpenELM with other small AI language models in a similar class, taken from the OpenELM research paper by Apple.

Enlarge / An table comparing OpenELM with other small AI language models in a similar class, taken from the OpenELM research paper by Apple.

Apple

Apple also released the code for CoreNet, a library it used to train OpenELM—and it also included reproducible training recipes that allow the weights (neural network files) to be replicated, which is unusual for a major tech company so far. As Apple says in its OpenELM paper abstract, transparency is a key goal for the company: “The reproducibility and transparency of large language models are crucial for advancing open research, ensuring the trustworthiness of results, and enabling investigations into data and model biases, as well as potential risks.”

By releasing the source code, model weights, and training materials, Apple says it aims to “empower and enrich the open research community.” However, it also cautions that since the models were trained on publicly sourced datasets, “there exists the possibility of these models producing outputs that are inaccurate, harmful, biased, or objectionable in response to user prompts.”

While Apple has not yet integrated this new wave of AI language model capabilities into its consumer devices, the upcoming iOS 18 update (expected to be revealed in June at WWDC) is rumored to include new AI features that utilize on-device processing to ensure user privacy—though the company may potentially hire Google or OpenAI to handle more complex, off-device AI processing to give Siri a long-overdue boost.

Apple releases eight small AI language models aimed at on-device use Read More »

microsoft’s-phi-3-shows-the-surprising-power-of-small,-locally-run-ai-language-models

Microsoft’s Phi-3 shows the surprising power of small, locally run AI language models

AI, Biz & IT, chatgpt, chatgtp, large language models, machine learning, microsoft, Microsoft Phi, Mistral, openai, Phi-2, Phi-3, Simon Willison, small language models / /

small packages —

Microsoft’s 3.8B parameter Phi-3 may rival GPT-3.5, signaling a new era of “small language models.”

–

An illustration of lots of information being compressed into a smartphone with a funnel.

Getty Images

On Tuesday, Microsoft announced a new, freely available lightweight AI language model named Phi-3-mini, which is simpler and less expensive to operate than traditional large language models (LLMs) like OpenAI’s GPT-4 Turbo. Its small size is ideal for running locally, which could bring an AI model of similar capability to the free version of ChatGPT to a smartphone without needing an Internet connection to run it.

The AI field typically measures AI language model size by parameter count. Parameters are numerical values in a neural network that determine how the language model processes and generates text. They are learned during training on large datasets and essentially encode the model’s knowledge into quantified form. More parameters generally allow the model to capture more nuanced and complex language-generation capabilities but also require more computational resources to train and run.

Some of the largest language models today, like Google’s PaLM 2, have hundreds of billions of parameters. OpenAI’s GPT-4 is rumored to have over a trillion parameters but spread over eight 220-billion parameter models in a mixture-of-experts configuration. Both models require heavy-duty data center GPUs (and supporting systems) to run properly.

In contrast, Microsoft aimed small with Phi-3-mini, which contains only 3.8 billion parameters and was trained on 3.3 trillion tokens. That makes it ideal to run on consumer GPU or AI-acceleration hardware that can be found in smartphones and laptops. It’s a follow-up of two previous small language models from Microsoft: Phi-2, released in December, and Phi-1, released in June 2023.

A chart provided by Microsoft showing Phi-3 performance on various benchmarks.

Enlarge / A chart provided by Microsoft showing Phi-3 performance on various benchmarks.

Phi-3-mini features a 4,000-token context window, but Microsoft also introduced a 128K-token version called “phi-3-mini-128K.” Microsoft has also created 7-billion and 14-billion parameter versions of Phi-3 that it plans to release later that it claims are “significantly more capable” than phi-3-mini.

Microsoft says that Phi-3 features overall performance that “rivals that of models such as Mixtral 8x7B and GPT-3.5,” as detailed in a paper titled “Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone.” Mixtral 8x7B, from French AI company Mistral, utilizes a mixture-of-experts model, and GPT-3.5 powers the free version of ChatGPT.

“[Phi-3] looks like it’s going to be a shockingly good small model if their benchmarks are reflective of what it can actually do,” said AI researcher Simon Willison in an interview with Ars. Shortly after providing that quote, Willison downloaded Phi-3 to his Macbook laptop locally and said, “I got it working, and it’s GOOD” in a text message sent to Ars.

A screenshot of Phi-3-mini running locally on Simon Willison's Macbook.

Enlarge / A screenshot of Phi-3-mini running locally on Simon Willison’s Macbook.

Simon Willison

“Most models that run on a local device still need hefty hardware,” says Willison. “Phi-3-mini runs comfortably with less than 8GB of RAM, and can churn out tokens at a reasonable speed even on just a regular CPU. It’s licensed MIT and should work well on a $55 Raspberry Pi—and the quality of results I’ve seen from it so far are comparable to models 4x larger.“

How did Microsoft cram a capability potentially similar to GPT-3.5, which has at least 175 billion parameters, into such a small model? Its researchers found the answer by using carefully curated, high-quality training data they initially pulled from textbooks. “The innovation lies entirely in our dataset for training, a scaled-up version of the one used for phi-2, composed of heavily filtered web data and synthetic data,” writes Microsoft. “The model is also further aligned for robustness, safety, and chat format.”

Much has been written about the potential environmental impact of AI models and datacenters themselves, including on Ars. With new techniques and research, it’s possible that machine learning experts may continue to increase the capability of smaller AI models, replacing the need for larger ones—at least for everyday tasks. That would theoretically not only save money in the long run but also require far less energy in aggregate, dramatically decreasing AI’s environmental footprint. AI models like Phi-3 may be a step toward that future if the benchmark results hold up to scrutiny.

Phi-3 is immediately available on Microsoft’s cloud service platform Azure, as well as through partnerships with machine learning model platform Hugging Face and Ollama, a framework that allows models to run locally on Macs and PCs.

Microsoft’s Phi-3 shows the surprising power of small, locally run AI language models Read More »

windows-vulnerability-reported-by-the-nsa-exploited-to-install-russian-malware

Windows vulnerability reported by the NSA exploited to install Russian malware

Biz & IT, exploits, microsoft, Security, vulnerabilities, Windows / /
Windows vulnerability reported by the NSA exploited to install Russian malware

Getty Images

Kremlin-backed hackers have been exploiting a critical Microsoft vulnerability for four years in attacks that targeted a vast array of organizations with a previously undocumented tool, the software maker disclosed Monday.

When Microsoft patched the vulnerability in October 2022—at least two years after it came under attack by the Russian hackers—the company made no mention that it was under active exploitation. As of publication, the company’s advisory still made no mention of the in-the-wild targeting. Windows users frequently prioritize the installation of patches based on whether a vulnerability is likely to be exploited in real-world attacks.

Exploiting CVE-2022-38028, as the vulnerability is tracked, allows attackers to gain system privileges, the highest available in Windows, when combined with a separate exploit. Exploiting the flaw, which carries a 7.8 severity rating out of a possible 10, requires low existing privileges and little complexity. It resides in the Windows print spooler, a printer-management component that has harbored previous critical zero-days. Microsoft said at the time that it learned of the vulnerability from the US National Security Agency.

On Monday, Microsoft revealed that a hacking group tracked under the name Forest Blizzard has been exploiting CVE-2022-38028 since at least June 2020—and possibly as early as April 2019. The threat group—which is also tracked under names including APT28, Sednit, Sofacy, GRU Unit 26165, and Fancy Bear—has been linked by the US and the UK governments to Unit 26165 of the Main Intelligence Directorate, a Russian military intelligence arm better known as the GRU. Forest Blizzard focuses on intelligence gathering through the hacking of a wide array of organizations, mainly in the US, Europe, and the Middle East.

Since as early as April 2019, Forest Blizzard has been exploiting CVE-2022-38028 in attacks that, once system privileges are acquired, use a previously undocumented tool that Microsoft calls GooseEgg. The post-exploitation malware elevates privileges within a compromised system and goes on to provide a simple interface for installing additional pieces of malware that also run with system privileges. This additional malware, which includes credential stealers and tools for moving laterally through a compromised network, can be customized for each target.

“While a simple launcher application, GooseEgg is capable of spawning other applications specified at the command line with elevated permissions, allowing threat actors to support any follow-on objectives such as remote code execution, installing a backdoor, and moving laterally through compromised networks,” Microsoft officials wrote.

GooseEgg is typically installed using a simple batch script, which is executed following the successful exploitation of CVE-2022-38028 or another vulnerability, such as CVE-2023-23397, which Monday’s advisory said has also been exploited by Forest Blizzard. The script is responsible for installing the GooseEgg binary, often named justice.exe or DefragmentSrv.exe, then ensuring that they run each time the infected machine is rebooted.

Windows vulnerability reported by the NSA exploited to install Russian malware Read More »

after-48-years,-zilog-is-killing-the-classic-standalone-z80-microprocessor-chip

After 48 years, Zilog is killing the classic standalone Z80 microprocessor chip

CP/M, microprocessors, microsoft, nintendo game boy, retrotech, Tech, TI-81, ti-83, Z80, Zilog, zx spectrum / /

rest in silicon —

Z80 powered game consoles, ZX Spectrum, Pac-Man, and a 1970s PC standard based on CP/M.

–

A cropped portion of a ca. 1980 ad for the Microsoft Z80 SoftCard, which allowed Apple II users to run the CP/M operating system.

Enlarge / A cropped portion of a ca. 1980 ad for the Microsoft Z80 SoftCard, which allowed Apple II users to run the CP/M operating system.

Microsoft

Last week, chip manufacturer Zilog announced that after 48 years on the market, its line of standalone DIP (dual inline package) Z80 CPUs is coming to an end, ceasing sales on June 14, 2024. The 8-bit Z80 architecture debuted in 1976 and powered a small-business-PC revolution in conjunction with CP/M, also serving as the heart of the Nintendo Game Boy, Sinclair ZX Spectrum, the Radio Shack TRS-80, the Pac-Man arcade game, and the TI-83 graphing calculator in various forms.

In a letter to customers dated April 15, 2024, Zilog wrote, “Please be advised that our Wafer Foundry Manufacturer will be discontinuing support for the Z80 product and other product lines. Refer to the attached list of the Z84C00 Z80 products affected.”

Designers typically use the Z84C00 chips because of familiarity with the Z80 architecture or to allow legacy system upgrades without needing significant system redesigns. And while many other embedded chip architectures have superseded these Z80 chips in speed, processing power, and capability, they remained go-to solutions for decades in products that didn’t need any extra horsepower.

Zilog will continue to manufacture the eZ80 microcontroller family, which was introduced in 2001 as a faster version of the Z80 series and comes in different physical package configurations (pin layouts).

Powering a microcomputer revolution

The 8-bit Z80 microprocessor was designed in 1974 by Federico Faggin as a binary-compatible, improved version of the Intel 8080 with a higher clock speed, a built-in DRAM refresh controller, and an extended instruction set. It was extensively used in desktop computers of the late 1970s and early 1980s, arcade video game machines, and embedded systems, and it became a cornerstone of several gaming consoles, like the Sega Master System.

The Tandy Radio Shack TRS-80 (1977), which used the Zilog Z80.

Enlarge / The Tandy Radio Shack TRS-80 (1977), which used the Zilog Z80.

SSPL/Getty Images

During the mid-late 1970s, the Z80 became a popular CPU for S-100 bus machines, which were early personal computers with a 100-pin modular bus system that allowed swapping cards to build systems based on parts from various manufacturers. Digital Research targeted the Z80 as a key platform for its CP/M operating system, and the association between Z80 and CP/M stuck, powering dozens of small business computers until the mid-1980s, when IBM PC clones running Microsoft’s MS-DOS became the new industry standard.

Interestingly, Microsoft’s first hardware product, the Z80 SoftCard for the Apple II in 1980, added the famous Zilog CPU to the classic personal computer and allowed users to run CP/M on that machine. In 1982, Bill Gates claimed that SoftCard installations represented the largest single user base of CP/M machines.

Last call in June 2024

Zilog is notably discontinuing several Z84C00 chips that are still available in classic 40-pin DIP packages resembling the classic Z80 CPU chips of the 1970s. (These standalone chips include a CPU and nothing else, unlike a microcontroller, which can include RAM and other accessory devices.) The DIP design features two rows of 20 pins with a plastic package in between that contains the actual embedded silicon chip, resembling the classic Z80 CPU chips of the 1970s.

After June 14, Zilog will stop taking orders, manufacture whatever orders are available if they are sufficient in quantity, then ship the last runs of the chips to resellers like Mouser Electronics and Digikey.

A classic dual inline package (DIP) version of the Z80 from the 1970s. It features two rows of 20 pins in a ceramic package.

Enlarge / A classic dual inline package (DIP) version of the Z80 from the 1970s. It features two rows of 20 pins in a ceramic package.

The discontinuation list provided by Zilog in its letter includes 13 products from the Z84C00 series, which are chips in the Z80 family that run at clock speeds from 6 to 20 MHz and maintain compatibility with the original Z80 architecture. Here’s the full list of part numbers that will be discontinued:

  • Z84C0006VEG
  • Z84C0006PEG
  • Z84C0010PEG
  • Z84C0008AEG
  • Z84C0020VEG
  • Z84C0008PEG
  • Z84C0010AEG
  • Z84C0008VEG
  • Z84C0010VEG
  • Z84C0010VEG00TR
  • Z84C0020AEG
  • Z84C0020PEG
  • Z84C0006AEG

So while the Z80 architecture will stick around in eZ80 form, it appears that this is the last call for newly manufactured standalone 8-bit Z80 CPU chips in the classic DIP form factor. We reached out to Zilog for clarification about its plans for the future of the Z80 platform but did not receive a response by press time.

After 48 years, Zilog is killing the classic standalone Z80 microprocessor chip Read More »

microsoft’s-vasa-1-can-deepfake-a-person-with-one-photo-and-one-audio-track

Microsoft’s VASA-1 can deepfake a person with one photo and one audio track

AI, AI ethics, AI safety, Biz & IT, Deepfakes, image synthesis, machine learning, microsoft, VASA-1, video synthesis / /

pics and it didn’t happen —

YouTube videos of 6K celebrities helped train AI model to animate photos in real time.

–

A sample image from Microsoft for

Enlarge / A sample image from Microsoft for “VASA-1: Lifelike Audio-Driven Talking Faces Generated in Real Time.”

On Tuesday, Microsoft Research Asia unveiled VASA-1, an AI model that can create a synchronized animated video of a person talking or singing from a single photo and an existing audio track. In the future, it could power virtual avatars that render locally and don’t require video feeds—or allow anyone with similar tools to take a photo of a person found online and make them appear to say whatever they want.

“It paves the way for real-time engagements with lifelike avatars that emulate human conversational behaviors,” reads the abstract of the accompanying research paper titled, “VASA-1: Lifelike Audio-Driven Talking Faces Generated in Real Time.” It’s the work of Sicheng Xu, Guojun Chen, Yu-Xiao Guo, Jiaolong Yang, Chong Li, Zhenyu Zang, Yizhong Zhang, Xin Tong, and Baining Guo.

The VASA framework (short for “Visual Affective Skills Animator”) uses machine learning to analyze a static image along with a speech audio clip. It is then able to generate a realistic video with precise facial expressions, head movements, and lip-syncing to the audio. It does not clone or simulate voices (like other Microsoft research) but relies on an existing audio input that could be specially recorded or spoken for a particular purpose.

Microsoft claims the model significantly outperforms previous speech animation methods in terms of realism, expressiveness, and efficiency. To our eyes, it does seem like an improvement over single-image animating models that have come before.

AI research efforts to animate a single photo of a person or character extend back at least a few years, but more recently, researchers have been working on automatically synchronizing a generated video to an audio track. In February, an AI model called EMO: Emote Portrait Alive from Alibaba’s Institute for Intelligent Computing research group made waves with a similar approach to VASA-1 that can automatically sync an animated photo to a provided audio track (they call it “Audio2Video”).

Trained on YouTube clips

Microsoft Researchers trained VASA-1 on the VoxCeleb2 dataset created in 2018 by three researchers from the University of Oxford. That dataset contains “over 1 million utterances for 6,112 celebrities,” according to the VoxCeleb2 website, extracted from videos uploaded to YouTube. VASA-1 can reportedly generate videos of 512×512 pixel resolution at up to 40 frames per second with minimal latency, which means it could potentially be used for realtime applications like video conferencing.

To show off the model, Microsoft created a VASA-1 research page featuring many sample videos of the tool in action, including people singing and speaking in sync with pre-recorded audio tracks. They show how the model can be controlled to express different moods or change its eye gaze. The examples also include some more fanciful generations, such as Mona Lisa rapping to an audio track of Anne Hathaway performing a “Paparazzi” song on Conan O’Brien.

The researchers say that, for privacy reasons, each example photo on their page was AI-generated by StyleGAN2 or DALL-E 3 (aside from the Mona Lisa). But it’s obvious that the technique could equally apply to photos of real people as well, although it’s likely that it will work better if a person appears similar to a celebrity present in the training dataset. Still, the researchers say that deepfaking real humans is not their intention.

“We are exploring visual affective skill generation for virtual, interactive charactors [sic], NOT impersonating any person in the real world. This is only a research demonstration and there’s no product or API release plan,” reads the site.

While the Microsoft researchers tout potential positive applications like enhancing educational equity, improving accessibility, and providing therapeutic companionship, the technology could also easily be misused. For example, it could allow people to fake video chats, make real people appear to say things they never actually said (especially when paired with a cloned voice track), or allow harassment from a single social media photo.

Right now, the generated video still looks imperfect in some ways, but it could be fairly convincing for some people if they did not know to expect an AI-generated animation. The researchers say they are aware of this, which is why they are not openly releasing the code that powers the model.

“We are opposed to any behavior to create misleading or harmful contents of real persons, and are interested in applying our technique for advancing forgery detection,” write the researchers. “Currently, the videos generated by this method still contain identifiable artifacts, and the numerical analysis shows that there’s still a gap to achieve the authenticity of real videos.”

VASA-1 is only a research demonstration, but Microsoft is far from the only group developing similar technology. If the recent history of generative AI is any guide, it’s potentially only a matter of time before similar technology becomes open source and freely available—and they will very likely continue to improve in realism over time.

Microsoft’s VASA-1 can deepfake a person with one photo and one audio track Read More »

llms-keep-leaping-with-llama-3,-meta’s-newest-open-weights-ai-model

LLMs keep leaping with Llama 3, Meta’s newest open-weights AI model

AI, AI assistants, Biz & IT, chatgpt, chatgtp, copilot, Google Gemini, LLaMA, Llama 2, Llama 3, machine learning, Meta, Meta AI Assistant, microsoft, Microsoft Copilot, Open Source, open weights, openai / /

computer-powered word generator —

Zuckerberg says new AI model “was still learning” when Meta stopped training.

–

A group of pink llamas on a pixelated background.

On Thursday, Meta unveiled early versions of its Llama 3 open-weights AI model that can be used to power text composition, code generation, or chatbots. It also announced that its Meta AI Assistant is now available on a website and is going to be integrated into its major social media apps, intensifying the company’s efforts to position its products against other AI assistants like OpenAI’s ChatGPT, Microsoft’s Copilot, and Google’s Gemini.

Like its predecessor, Llama 2, Llama 3 is notable for being a freely available, open-weights large language model (LLM) provided by a major AI company. Llama 3 technically does not quality as “open source” because that term has a specific meaning in software (as we have mentioned in other coverage), and the industry has not yet settled on terminology for AI model releases that ship either code or weights with restrictions (you can read Llama 3’s license here) or that ship without providing training data. We typically call these releases “open weights” instead.

At the moment, Llama 3 is available in two parameter sizes: 8 billion (8B) and 70 billion (70B), both of which are available as free downloads through Meta’s website with a sign-up. Llama 3 comes in two versions: pre-trained (basically the raw, next-token-prediction model) and instruction-tuned (fine-tuned to follow user instructions). Each has a 8,192 token context limit.

A screenshot of the Meta AI Assistant website on April 18, 2024.

Enlarge / A screenshot of the Meta AI Assistant website on April 18, 2024.

Benj Edwards

Meta trained both models on two custom-built, 24,000-GPU clusters. In a podcast interview with Dwarkesh Patel, Meta CEO Mark Zuckerberg said that the company trained the 70B model with around 15 trillion tokens of data. Throughout the process, the model never reached “saturation” (that is, it never hit a wall in terms of capability increases). Eventually, Meta pulled the plug and moved on to training other models.

“I guess our prediction going in was that it was going to asymptote more, but even by the end it was still leaning. We probably could have fed it more tokens, and it would have gotten somewhat better,” Zuckerberg said on the podcast.

Meta also announced that it is currently training a 400B parameter version of Llama 3, which some experts like Nvidia’s Jim Fan think may perform in the same league as GPT-4 Turbo, Claude 3 Opus, and Gemini Ultra on benchmarks like MMLU, GPQA, HumanEval, and MATH.

Speaking of benchmarks, we have devoted many words in the past to explaining how frustratingly imprecise benchmarks can be when applied to large language models due to issues like training contamination (that is, including benchmark test questions in the training dataset), cherry-picking on the part of vendors, and an inability to capture AI’s general usefulness in an interactive session with chat-tuned models.

But, as expected, Meta provided some benchmarks for Llama 3 that list results from MMLU (undergraduate level knowledge), GSM-8K (grade-school math), HumanEval (coding), GPQA (graduate-level questions), and MATH (math word problems). These show the 8B model performing well compared to open-weights models like Google’s Gemma 7B and Mistral 7B Instruct, and the 70B model also held its own against Gemini Pro 1.5 and Claude 3 Sonnet.

A chart of instruction-tuned Llama 3 8B and 70B benchmarks provided by Meta.

Enlarge / A chart of instruction-tuned Llama 3 8B and 70B benchmarks provided by Meta.

Meta says that the Llama 3 model has been enhanced with capabilities to understand coding (like Llama 2) and, for the first time, has been trained with both images and text—though it currently outputs only text. According to Reuters, Meta Chief Product Officer Chris Cox noted in an interview that more complex processing abilities (like executing multi-step plans) are expected in future updates to Llama 3, which will also support multimodal outputs—that is, both text and images.

Meta plans to host the Llama 3 models on a range of cloud platforms, making them accessible through AWS, Databricks, Google Cloud, and other major providers.

Also on Thursday, Meta announced that Llama 3 will become the new basis of the Meta AI virtual assistant, which the company first announced in September. The assistant will appear prominently in search features for Facebook, Instagram, WhatsApp, Messenger, and the aforementioned dedicated website that features a design similar to ChatGPT, including the ability to generate images in the same interface. The company also announced a partnership with Google to integrate real-time search results into the Meta AI assistant, adding to an existing partnership with Microsoft’s Bing.

LLMs keep leaping with Llama 3, Meta’s newest open-weights AI model Read More »

german-state-gov.-ditching-windows-for-linux,-30k-workers-migrating

German state gov. ditching Windows for Linux, 30K workers migrating

Biz & IT, germany, LibreOffice, Linux, microsoft, Software, Windows / /

Open source FTW —

Schleswig-Holstein looks to succeed where Munich failed.

–

many penguins

Schleswig-Holstein, one of Germany’s 16 states, on Wednesday confirmed plans to move tens of thousands of systems from Microsoft Windows to Linux. The announcement follows previously established plans to migrate the state government off Microsoft Office in favor of open source LibreOffice.

As spotted by The Document Foundation, the government has apparently finished its pilot run of LibreOffice and is now announcing plans to expand to more open source offerings.

In 2021, the state government announced plans to move 25,000 computers to LibreOffice by 2026. At the time, Schleswig-Holstein said it had already been testing LibreOffice for two years.

As announced on Minister-President Daniel Gunther’s webpage this week, the state government confirmed that it’s moving all systems to the Linux operating system (OS), too. Per a website-provided translation:

With the cabinet decision, the state government has made the concrete beginning of the switch away from proprietary software and towards free, open-source systems and digitally sovereign IT workplaces for the state administration’s approximately 30,000 employees.

The state government is offering a training program that it said it will update as necessary.

Regarding LibreOffice, the government maintains the possibility that some jobs may use software so specialized that they won’t be able to move to open source software.

In 2021, Jan Philipp Albrecht, then-minister for Energy, Agriculture, the Environment, Nature, and Digitalization of Schleswig-Holstein, discussed interest in moving the state government off of Windows.

“Due to the high hardware requirements of Windows 11, we would have a problem with older computers. With Linux we don’t have that,” Albrecht told Heise magazine, per a Google translation.

This week’s announcement also said that the Schleswig-Holstein government will ditch Microsoft Sharepoint and Exchange/Outlook in favor of open source offerings Nextcloud and Open-Xchange, and Mozilla Thunderbird in conjunction with the Univention active directory connector.

Schleswig-Holstein is also developing an open source directory service to replace Microsoft’s Active Directory and an open source telephony offering.

Digital sovereignty dreams

Explaining the decision, the Schleswig-Holstein government’s announcement named enhanced IT security, cost efficiencies, and collaboration between different systems as its perceived benefits of switching to open source software.

Further, the government is pushing the idea of digital sovereignty, with Schleswig-Holstein Digitalization Minister Dirk Schrödter quoted in the announcement as comparing the concept’s value to that of energy sovereignty. The announcement also quoted Schrödter as saying that digital sovereignty isn’t achievable “with the current standard IT workplace products.”

Schrödter pointed to the state government’s growing reliance on cloud services and said that with related proprietary software, users have no influence on data flow and whether that data makes its way to other countries.

Schrödter also claimed that the move would help with the state’s budget by diverting money from licensing fees to “real programming services from our domestic digital economy” that could also create local jobs.

In 2021, Albrecht said the state was reaching its limits with proprietary software contracts because “license fees have continued to rise in recent years,” per Google’s translation.

“Secondly, regarding our goals for the digitalization of administration, open source simply offers us more flexibility,” he added.

At the time, Albrecht claimed that 90 percent of video conferences in the state government ran on the open source program Jitsi, which was advantageous during the COVID-19 pandemic because the state was able to quickly increase video conferencing capacity.

Additionally, he said that because the school portal was based on (unnamed) open source software, “we can design the interface flexibly and combine services the way we want.”

There are numerous other examples globally of government entities switching to Linux in favor of open source technology. Federal governments with particular interest in avoiding US-based technologies, including North Korea and China, are some examples. The South Korean government has also shared plans to move to Linux by 2026, and the city of Barcelona shared migration plans in 2018.

But some government bodies that have made the move regretted it and ended up crawling back to Windows. Vienna released the Debian-based distribution WIENUX in 2005 but gave up on migration by 2009.

In 2003, Munich announced it would be moving some 14,000 PCs off Windows and to Linux. In 2013, the LiMux project finished, but high associated costs and user dissatisfaction resulted in Munich announcing in 2017 that it would spend the next three years reverting back to Windows.

Albrecht in 2021 addressed this failure when speaking to Heise, saying, per Google’s translation:

The main problem there was that the employees weren’t sufficiently involved. We do that better. We are planning long transition phases with parallel use. And we are introducing open source step by step where the departments are ready for it. This also creates the reason for further rollout because people see that it works.

German state gov. ditching Windows for Linux, 30K workers migrating Read More »

copilot-key-is-based-on-a-button-you-probably-haven’t-seen-since-ibm’s-model-m

Copilot key is based on a button you probably haven’t seen since IBM’s Model M

AI, copilot, generative ai, Keyboards, laptops, microsoft, Tech, Windows, Windows 11 / /

Microsoft chatbot button —

Left-Shift + Windows key + F23

–

A Dell XPS 14 laptop with a Copilot key.

Enlarge / A Dell XPS 14 laptop. The Copilot key is to the right of the right-Alt button.

In January, Microsoft introduced a new key to Windows PC keyboards for the first time in 30 years. The Copilot key, dedicated to launching Microsoft’s eponymous generative AI assistant, is already on some Windows laptops released this year. On Monday, Tom’s Hardware dug into the new addition and determined exactly what pressing the button does, which is actually pretty simple. Pushing a computer’s integrated Copilot button is like pressing left-Shift + Windows key + F23 simultaneously.

Tom’s Hardware confirmed this after wondering if the Copilot key introduced a new scan code to Windows or if it worked differently. Using the scripting program AuthoHotkey with a new laptop with a Copilot button, Tom’s Hardware discovered the keystrokes registered when a user presses the Copilot key. The publication confirmed with Dell that “this key assignment is standard for the Copilot key and done at Microsoft’s direction.”

F23

Surprising to see in that string of keys is F23. Having a computer keyboard with a function row or rows that take you from F1 all the way to F23 is quite rare today. When I try to imagine a keyboard that comes with an F23 button, vintage keyboards come to mind, more specifically buckling spring keyboards from IBM.

IBM’s Model F, which debuted in 1981 and used buckling spring switches over a capacitive PCB, and the Model M, which launched in 1985 and used buckling spring switches over a membrane sheet, both offered layouts with 122 keys. These layouts included not one, but two rows of function keys that would leave today’s 60 percent keyboard fans sweating over the wasted space.

But having 122 keys was helpful for keyboards tied to IBM business terminals. The keyboard layout even included a bank of keys to the left of the primary alpha block of keys for even more forms of input.

An IBM Model M keyboard with an F23 key.

Enlarge / An IBM Model M keyboard with an F23 key.

The 122-key keyboard layout with F23 lives on. Beyond people who still swear by old Model F and M keyboards, Model F Labs and Unicomp both currently sell modern buckling spring keyboards with built-in F23 buttons. Another reason a modern Windows PC user might have access to an F23 key is if they use a macro pad.

But even with those uses in mind, the F23 key remains rare. That helps explain why Microsoft would use the key for launching Copilot; users are unlikely to have F23 programmed for other functions. This was also likely less work than making a key with an entirely new scan code.

The Copilot button is reprogrammable

When I previewed Dell’s 2024 XPS laptops, a Dell representative told me that the integrated Copilot key wasn’t reprogrammable. However, in addition to providing some interesting information about the newest PC key since the Windows button, Tom’s Hardware’s revelation shows why the Copilot key is actually reprogrammable, even if OEMs don’t give users a way to do so out of the box. (If you need help, check out the website’s tutorial for reprogramming the Windows Copilot key.)

I suspect there’s a strong interest in reprogramming that button. For one, generative AI, despite all its hype and potential, is still an emerging technology. Many don’t need or want access to any chatbot—let alone Microsoft’s—instantly or even at all. Those who don’t use their system with a Microsoft account have no use for the button, since being logged in to a Microsoft account is required for the button to launch Copilot.

A rendering of the Copilot button.

Enlarge / A rendering of the Copilot button.

Microsoft

Additionally, there are other easy ways to launch Copilot on a computer that has the program downloaded, like double-clicking an icon or pressing Windows + C, that make a dedicated button unnecessary. (Ars Technica asked Microsoft why the Copilot key doesn’t just register Windows + C, but the company declined to comment. Windows + C has launched other apps in the past, including Cortana, so it’s possible that Microsoft wanted to avoid the Copilot key performing a different function when pressed on computers that use Windows images without Copilot.)

In general, shoehorning the Copilot key into Windows laptops seems premature. Copilot is young and still a preview; just a few months ago, it was called Bing Chat. Further, the future of generative AI, including its popularity and top uses, is still forming and could evolve substantially during the lifetime of a Windows laptop. Microsoft’s generative AI efforts could also flounder over the years. Imagine if Microsoft went all-in on Bing back in the day and made all Windows keyboards have a Bing button, for example. Just because Microsoft wants something to become mainstream doesn’t mean that it will.

This all has made the Copilot button seem more like a way to force the adoption of Microsoft’s chatbot than a way to improve Windows keyboards. Microsoft has also made the Copilot button a requirement for its AI PC certification (which also requires an integrated neural processing unit and having Copilot pre-installed). Microsoft plans to make Copilot keys a requirement for Windows 11 OEM PCs eventually, it told Ars Technica in January.

At least for now, the basic way that the Copilot button works means you can turn the key into something more useful. Now, the tricky part would be finding a replacement keycap to eradicate Copilot’s influence from your keyboard.

Listing image by Microsoft

Copilot key is based on a button you probably haven’t seen since IBM’s Model M Read More »

microsoft-splits-up-the-teams-and-office-apps-worldwide,-following-eu-split

Microsoft splits up the Teams and Office apps worldwide, following EU split

Biz & IT, microsoft, microsoft 365, Microsoft office, Microsoft Teams, office, Teams / /

different teams —

Changes may save a bit of money for people who want Office apps without Teams.

– Updated

Teams is being decoupled from the other Office apps worldwide, six months after Microsoft did the same thing for the EU.

Enlarge / Teams is being decoupled from the other Office apps worldwide, six months after Microsoft did the same thing for the EU.

Microsoft/Andrew Cunningham

Months after unbundling the apps in the European Union, Microsoft is taking the Office and Teams breakup worldwide. Reuters reports that Microsoft will begin selling Teams and the other Microsoft 365 apps to new commercial customers as separate products with separate price tags beginning today.

“To ensure clarity for our customers, we are extending the steps we took last year to unbundle Teams from M365 and O365 in the European Economic Area and Switzerland to customers globally,” a Microsoft spokesperson told Ars. “Doing so also addresses feedback from the European Commission by providing multinational companies more flexibility when they want to standardize their purchasing across geographies.”

The unbundling is a win for other team communication apps like Slack and videoconferencing apps like Zoom, both of which predate Teams but haven’t had the benefits of the Office apps’ huge established user base.

The separation follows an EU regulatory investigation that started in July of 2023, almost exactly three years after Slack initially filed a complaint alleging that Microsoft was “abusing its market dominance to extinguish competition in breach of European Union competition law.”

In August of 2023, Microsoft announced that it would be unbundling the apps in the EU and Switzerland in October. Bloomberg reported in September that Zoom had met with EU and US Federal Trade Commission regulators about Microsoft, further ratcheting up regulatory pressure on Microsoft.

In October, Microsoft European Government Affairs VP Nanna-Louise Linde described the unbundling and other moves as “proactive changes that we hope will start to address these concerns in a meaningful way,” though the EU investigation is ongoing, and the company may yet be fined. Linde also wrote that Microsoft would allow third-party apps like Zoom and Slack to integrate more deeply with the Office apps and that it would “enable third-party solutions to host Office web applications.”

Microsoft has put up a blog post detailing its new pricing structure here—for now, the changes only affect the Microsoft 365 plans for the Business, Enterprise, and Frontline versions of Microsoft 365. Consumer, Academic, US Government, and Nonprofit editions of Microsoft 365 aren’t changing today and will still bundle Teams as they did before.

Current Office/Microsoft 365 Enterprise customers who want to keep using the Office apps and Teams together can continue to subscribe to both at their current prices. New subscribers to the Enterprise versions of Microsoft 365/Office 365 can pay $5.25 per user per month for Teams, whether they’re buying Teams as standalone software or adding it on top of a Teams-free Office/Microsoft 365 subscription.

For the Business and Frontline Microsoft 365 versions, you can either buy the version with Teams included for the same price as before, or choose a new Teams-less option that will save you a couple of dollars per user per month. For example, the Teams-less version of Microsoft 365 Business Standard costs $10.25 per user per month, compared to $12.50 for the version that includes Teams.

Updated April 1, 2024, at 4: 12 pm to add more details about pricing and a link to Microsoft’s official blog post about the announcement; also added a statement from a Microsoft spokesperson.

Microsoft splits up the Teams and Office apps worldwide, following EU split Read More »