Month: July 2024

Apple is ready for government — and your business

As governments everywhere continue to deal with the economically damaging fallout of last week’s appalling Crowdstrike/Microsoft disaster, it’s no surprise to see more governments switching to Apple’s most stable platform.

To put things into context, Parametrix Insurance has analyzed the consequences of last week’s global blue screen of death incident and assesses the economic cost of the moment on the world economy as being in the multiple billions of dollars. Estimated direct financial loss across the Fortune 500 was $5.4 billion, Parametrix said. 

That’s a painful amount to lose.

This did real damage

While it has subsequently been claimed that 97% of affected systems are back in operation, that still leaves hundreds of thousands of business-critical PCs that aren’t doing their job. I’ll be looking more closely at these consequences next week. I imagine businesses that became reliant on these products are already exploring their options — after all, that kind of $5.4 billion cost was almost certainly never anticipated by IT departments when estimating the TCOs of their PC fleets.

These incidents will undoubtedly make a lot of people think twice when it comes to their next hardware upgrade. After all, as business becomes increasingly mobile, and services migrate to the cloud, the strength of vendor lock-in is shrinking to the extent that it is becoming much easier to transition to multiple platforms to build business resilience against future debacles — or move to a platform characterized by lack of such drama, which is Apple.

Is it time to migrate?

Apple is ready if you are. 

Not only is the company now equipped with a wide and diverse range of enterprise-focused companies such as Addigy, Jamf, Kandji and all the other firms I speak with each week, but its future-focused platforms are inherently more resilient by design. That’s why Apple dumped support for kernel extensions years ago, among other examples. While no platform is ever completely secure, the vast majority of problems on Apple’s platforms emerge through user error, not globally deployed automated PC-borking software updates.

Of course, the bias that Apple is a consumer product that isn’t fit for the enterprise runs deep, and shifting that view is taking time — though events such as the Crowdstrike disaster should help people question that opinion. 

I think the momentum to diversify is growing. 

Apple is ready for government

That the German government is switching to Apple speaks to this trend. The procurement office of Germany’s Federal Ministry of the Interior and Community’s (BMI) recently concluded a deal with Bechtle for the supply of up to 300,000 Apple devices across the entire federal administration. That means all government agencies will use iPhones and iPads equipped with Apple’s iNDIGO (iOS Native Devices in Government Operation) platform.

iNDIGO offers built-in security features, regular updates, hardware-based encryption and strict data protection policies and has been approved by the German Federal Office for Information Security (BSI) for the secure sharing of confidential content classified as “for official use only” (VS-NfD).” 

It’s very similar to the increasing adoption of Apple products across US government. Apple is also “actively participating in several governmental security evaluations and certifications globally,” the company says.

Time to switch?

As I see it, for companies looking to build more resilience in IT after the Crowdstrike strike, news that Apple’s platforms are considered good enough and secure enough for active use by national governments should lay to rest the utterly mistaken myth that Apple’s devices aren’t good for business. Apple is now ready for government. It’s ready for business.

Finally, of course, regardless of whatever platforms you choose to deploy, it seems sensible to check the terms and conditions of any mission-critical service to ensure you’ll get compensated in the event a software patch breaks your business and costs you money. Why should any customer pay for a product that fails?

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.

Google’s new AI system reaches a mathematical milestone

Google’s AI unit Deepmind has demonstrated two new AI systems, AlphaProof and AlphaGeometry 2, which can successfully solve complex mathematical problems, representing a significant milestone in AI development, Reuters reports .

Today’s AI models work by being able to recognize patterns and statistically predict, for example, the next word that should be generated. However, this is not enough to handle abstract mathematics, which instead requires a system that can reason in a way that more closely resembles human intelligence.

AlpaProof and AlphaGeometry 2 succeeded in solving four out of six questions in this year’s edition of the International Mathematical Olympiad (IMO). One question was solved in just a few minutes; others took up to three days to solve, which is longer than the competition allows. AlpaProof also managed to solve the competition’s most difficult problem, which only five of 600 human participants managed to do.

Open AI CEO: Authoritarian states want to take over the world with AI

Open AI CEO Sam Altman has written an attention-grabbing column in The Washington Post about the risks of artificial intelligence (AI), warning that authoritarian countries such as Russia and China want to control the development of AI systems — something that could eventually threaten humanity.

For example, Russian President Vladimir Putin has said that whoever wins the AI ​​battle will be the “ruler of the world”, while China has a stated goal of becoming the leader in AI by 2030. To prevent that from happening, the United States and its allies should work together to ensure that AI is used as a tool for freedom and democracy, Altman wrote.

To achieve that goal, substantial investment in security, infrastructure, diplomacy and dialogue will be required, he argued.

OpenAI’s AI-powered SearchGPT is set to challenge Google’s web search dominance

OpenAI has thrown down the gauntlet to search giants like Google and Bing with the launch of its AI-powered search engine, SearchGPT.

Currently, in testing with a small group of users and publishers, SearchGPT combines the advanced conversational abilities of OpenAI’s large language models with real-time web data to provide “fast, accurate answers complete with clear and relevant sources.”

“Getting answers on the web can take a lot of effort,” OpenAI said in a statement. “We believe that by enhancing the conversational capabilities of our models with real-time information from the web, finding what you’re looking for can be faster and easier,” the statement added.

OpenAI’s entry into the search market presents a significant challenge to traditional search giants including Google and Bing, as well as for AI-enabled chatbots such as Perplexity.

Microsoft is already using OpenAI’s tools to add AI-driven capabilities to its search. Google also announced similar features for its search engine in May.

However, unlike conventional search engines, which often require users to sift through multiple pages of results, SearchGPT claims to “respond to your questions with up-to-date information from the web while giving you clear links to relevant sources.”

“Google’s empire hinges on search but lacks a cleaner, contextual, and conversational experience,” said Neil Shah, VP for research and partner at Counterpoint Research. “This is where OpenAI could turn the industry upside down with growing usage and making SearchGPT the cornerstone for their future GPT models.”

This approach not only simplifies the search process but also enhances user engagement with content publishers, the statement added.

According to OpenAI users can ask follow-up questions, similar to a natural dialogue, with SearchGPT continuously building upon the context of their search.

“ChatGPT ushered in a paradigm shift from passive ‘search’ to active ‘seek’, revolutionizing how we interact with information,” said Prabhu Ram, VP for Industry Research Group at CyberMedia Research. “SearchGPT is OpenAI’s strategic next step, capitalizing on consumer enthusiasm for conversational AI and challenging industry giants like Google and emerging competitors like Perplexity.”

The potential impact extends beyond traditional search engines. For ChatGPT competitors, SearchGPT represents a significant step forward in AI-powered information retrieval. By integrating real-time web information with its existing capabilities, OpenAI could further solidify its position in the conversational AI landscape.

Addressing publishers’ concerns

One of the areas OpenAI’s SearchGPT has tried to address is how content from various publishers is acquired, and used and whether due credit has been given to content creators by the LLM model.

“For decades, search has been a foundational way for publishers and creators to reach users,” the statement added. “SearchGPT is designed to help users connect with publishers by prominently citing and linking to them in searches.”

“While SearchGPT has emphasized content licensing agreements, it will be intriguing to observe how it addresses legal or ethical challenges around content usage — issues that have notably affected Perplexity AI,” said Ram.

OpenAI has been accused of copyright violations by multiple leading global publications and a few publications have claimed millions of their articles were used to train ChatGPT without their permission. In June this year, Forbes accused Perplexity of using its proprietary content without seeking permission or giving due credit.

OpenAI has vowed not to repeat its mistakes and said it would “strive to maintain a healthy ecosystem for publishers and creators.” It would prominently display sources alongside search results, ensuring transparency and crediting the origin of information.

Additionally, publishers will now be able to manage how their content appears within SearchGPT, empowering them to have control over their online presence in this new search paradigm.

“We’re using AI to enhance this experience by highlighting high-quality content in a conversational interface with multiple opportunities for users to engage,” OpenAI added in the statement.

This focus on user experience and publisher control could contribute to a more diverse and dynamic search landscape.

“One thing I liked about Search GPT’s approach is they are trying to connect publishers and the readers and giving them more control on how they want to manage and showcase their content,” said Shah. “So, this is a great approach from OpenAI to empower publishers while they can remain autonomous but discoverable in a way they want to.”

It’s a big threat to Google’s business model,” he added.

However, not all publishers seem to be amused with the new search engine. “Given OpenAI’s reputation I don’t see content creators benefit much,” said Varun Krishnan, founder of Phone Arena, a technology publication based in India. “Both OpenAI and Perplexity have been accused of featuring the work of publishers without due credits. I wish they cite the source and give due credits to publishers as promised.”

Another area of concern is how publishers and content creators are going to recalibrate their content optimization going forward.

“From a content discovery point of view all content creators including those selling products and services have optimized their web presences for search engines,” said Faisal Kawoosa, founder and chief analyst at Techarc. “So, the SearchGPT will have to educate all of us on how it’s going to be different.”

How to repair Windows 10 or 11 in 4 steps

Sometimes a Windows 10 system starts misbehaving to the point where repair is needed. This often takes the form of worsening performance or stability, and can originate from damage to, loss of, or corruption of Windows system files (normally in the C:\Windows folder hierarchy). When that happens, users would be well advised to break out the following routine to help them set things back to rights.

This story provides a sequence of techniques to try if you’re frequently experiencing problems like slowdowns and crashes. All of these methods start with the assumption that you’re at least able to get into the Windows environment. If Windows won’t start up at all, see my story “What to do when Windows won’t boot,” which covers techniques such as WinRE Startup Repair, using a third-party boot repair tool, and more.

I initially wrote this story in 2016, but as Windows 10 has evolved and Windows 11 has emerged, some of these steps have changed a bit. I’ve now updated it for Windows 10 22H2 and Windows 11 23H2, the current versions. Most of the steps are the same in Windows 10 and 11, but as you’ll see, Microsoft has made Step 3 much simpler in Windows 11.

The Windows repair drill

In this exercise, the idea is to march through the steps in order. Thus, first try the initial step in the sequence. If that doesn’t fix what’s broken, advance to the next step. Keep working through these steps in order until you reach the end, and you are bound to fix the vast majority of problems. In fact, the only remaining step at that point would be to rebuild or replace the system on which Windows runs, and that’s outside the scope of this story.

The amount of time and effort required for each step goes up incrementally. Some steps involve additional work to get your PC more or less back to where it was prior to taking that step. Thus, the most important bit of advice I can dispense for those who must venture beyond Step 1 is this: make a complete backup of your system to provide a source for files and settings that might otherwise go missing. Ignore this at your own risk.

Back it up so you don’t lose it!

Any time you make major changes to a Windows system, such as installing a major application suite like Microsoft Office or performing an OS upgrade or repair, it’s a good idea to begin that process with a complete backup. The important thing is to use a backup tool that creates an image backup of your boot/system drive. That allows you to rewrite that drive and restore your system to normal operation should anything go wrong in your changes or repairs. See my article “How to make a Windows 10 or 11 image backup” for instructions.

Step 1: Using DISM and SFC for system repair

It’s now considered a best practice to run DISM before running the System File Checker (SFC). Earlier versions of this story did things in reverse order — but that was before I got direct advice from Microsoft to run DISM first, SFC second. Let me explain…

DISM is the Deployment Image Servicing and Management tool. Built into Windows, it’s a Swiss Army knife for working with Windows OS images, both offline and online. DISM can often fix things that SFC can detect but that it cannot itself fix. DISM supports a raft of capabilities with switches and parameters to match, but basic system file repair syntax is reasonably easy (though it often requires multiple passes before it sets things straight).

You must run DISM in an Administrator: Command Prompt window (press the Windows key + X combination and select Command Prompt (Admin) from the resulting pop-up menu) or in an Administrative PowerShell or Windows Terminal session. Here’s some sample syntax (consult the Microsoft Learn DISM reference for all the gory details):

dism /online /cleanup-image /checkhealth

Typically, when you run this command, the output will show nothing is amiss, as in Figure 1:

using the dism checkhealth command

Figure 1: DISM can check the health of any Windows image. Here, it reports A-OK for a production PC.

Ed Tittel / IDG

The /checkhealth command simply checks the running image (that’s what /online means) to see if it can detect any damage or corruption. Because it’s only checking file hashes and signatures, this command completes in under a minute on most machines. The good thing about /checkhealth is that it not only tells you if it finds damage but if that damage is repairable or not. If it’s repairable, you can proceed to the /restorehealth option (covered in the following paragraphs); if it’s non-repairable, jump to Step 2.

If DISM reports that component store corruption is found and is repairable, you should attempt repairs. That syntax reads:

dism /online /cleanup-image /restorehealth

Even if errors were detected in the preceding step, this command should complete successfully. It replaces corrupt or questionable elements in the Windows Side-by-Side (WinSxS) store, reading from a local copy of your Windows image files by default.

The /restorehealth command can be tricky to use. Because it actually repairs a Windows image, it needs a source from which to attempt such repairs. You can omit the /source option, but if you do, the command will try to grab its files from Windows Update over the internet. This may or may not work, depending on firewall settings on your network.

A safer bet is to point DISM at a known good source for Windows image components on the local machine (or on your local network). This can be a Windows image (.wim) file or a compressed Windows image (.esd, used for electronic software downloads of Windows installations, as the file extension indicates). You can also point to a separate WinSXS folder (the usual directory path is C:\Windows\WinSXS) taken from another PC with similar hardware for which dism … /checkhealth returns no issues.

The syntax for image files is where things can get interesting. To point to the install.wim file that shows up in ISO downloads for the Windows 10 or 11 installer on a USB flash drive designated L:, for example, you must use the following source specification (which uses the first image it finds inside the .wim file, designated :1):

/source:WIM:L:sourcesinstall.wim:1

To use an ESD file, change the wim items to esd, as shown here:

/source:ESD:L:sourcesinstall.esd:1

If you run into difficulties getting this to work, try adding the /limitaccess switch to your command strings. This stops DISM from attempting to grab source files from Windows Update, which it will sometimes do even if pointed at a local source. If you just can’t get either of these to work, try a WinSXS folder instead.

Note: the version and language for the /source image must be identical to that for the image to be repaired. This may mean using UUP Dump, Microsoft’s Media Creation Tool, the Windows Insiders page, or even Visual Studio Subscriptions (subscription required) to grab the correct Windows version/build. See my article “The best places to find Windows ISOs” for details on how to use these sources. It also explains how to create media and mount an ISO file to give DISM the right /source from which to work.

Figure 2 shows typical DISM output using /restorehealth:

using the dism restorehealth command

Figure 2: Even if /restorehealth doesn’t DO anything, it reports successful completion (unless there’s an error, in which case it may report either failure or success, depending on what happens).

Ed Tittel / IDG

Now it’s time to run SFC, Microsoft’s System File Checker. To quote Microsoft Learn, SFC “scans and verifies the versions of all protected system files.” Thus, it tells you if it finds anything amiss. Enter the command SFC /scannow at an Administrative Command Prompt or Administrative PowerShell session. The command takes a few minutes to complete, depending on your PC hardware. Unlike DISM, SFC scans and makes repairs in one go.

Figure 3 shows sample output when SFC finds files in need of repair. If this occurs on your PC, repeat the SFC /scannow command until you see the clean bill of health shown in Figure 4. This seldom takes more than two or three tries (on the system shown in Figures 3 and 4, it took one repeat only).

sfc scannow command finding and repairing corrupt files

Figure 3: When SFC finds errors, it fixes (and logs) them.

Ed Tittel / IDG

sfc scannow no violations found

Figure 4: Repeat SFC /scannow until you see “did not find any integrity violations.”

Ed Tittel / IDG

In my experience, most problems are solved at this point. Thus, the odds are favorable that you won’t have to continue onto Step 2. But if DISM can’t find your Windows image, or SFC can’t repair your system files, proceed to Step 2.

Step 2: Try a recent image backup

Restoring an image backup works like this: launch the backup/restore software, tell it to restore, choose the image (usually by date and time) to which you’d like to roll back, then initiate the restore procedure. For example, Macrium Reflect, my preferred image backup tool, offers a Restore menu, then allows you to browse stored backup files to pick the one you want in an Explorer-like Open window.

For step-by-step instructions for making and restoring image backups, see my story “How to make a Windows 10 or 11 image backup.”

If reverting to an image backup doesn’t work (or you don’t have one), it’s time to move on to Step 3.

Step 3: In-place upgrade to current Windows version

For Windows 10 or 11, this technique essentially involves overwriting the current OS installation with a fresh new copy while leaving data files, settings, and preferences alone. That means running the Windows 10 or 11 installer for the same version/build currently running on the machine from inside Windows itself. As you will learn at the end of this section, this is markedly easier for Windows 11 than it is for Windows 10.

Performing an in-place upgrade in Windows 10

To accomplish this for Windows 10, you’ll need installation media for the same edition (Home, Pro, Enterprise, or Education), language (en-US for American English, for example), and build as the currently installed edition. Bittedness must also be identical: 32-bit media for a 32-bit install, 64-bit media for a 64-bit install. (See The best places to find Windows ISOs for info on obtaining the correct Windows version/build from UUP Dump, or the Media Creation tool, or other sources. I recommend building an installable and bootable USB flash drive for this purpose; you may need it again sometime.

Before launching into this process, be sure to log in to Windows 10 with an administrative account. Most experts recommend that you disable or uninstall any third-party antivirus or security software (anything other than Windows Defender, in other words) and turn off Fast Boot and Secure Boot in your UEFI firmware settings (if turned on). You can reinstate these items when the in-place upgrade is complete.

Performing an in-place upgrade is easy. Run setup.exe from the USB flash drive, or from the mounted ISO, which should reside on some drive other than the system/boot device. The Windows installer will do the rest and will guide you through the installation process. On the “Ready to install” screen, be sure both “Install Windows 10” and “Keep personal files and apps” are checked. The second checkmark is particularly important, because that tells you it’s indeed an upgrade install, and leaves non-OS files and settings (mostly) alone.

The whole process takes 10 to 30 minutes, depending on your PC’s hardware. You’ll need to reestablish your network connection and set your time zone (unless you live in the Pacific time zone where Microsoft has its HQ). For more detailed instructions, including a list of cleanup tasks to perform after the in-place install, see “How to fix Windows 10 with an in-place upgrade install.”

Performing an in-place upgrade in Windows 11

If you’re running Windows 11 and want to perform an in-place upgrade install, that process is dead easy. Click Start > Settings > System > Recovery (four clicks). As shown in Figure 5, you need to click only the Reinstall now button under the “Fix problems using Windows Update” heading to make this happen.

windows 11 system recovery screen with reinstall now button

Figure 5: Once you navigate to this screen, click the “Reinstall now” button. Done!

Ed Tittel / IDG

Windows Update does the rest: it downloads all the necessary files and builds a compatible repair image, no additional effort required. IMO, it’s one of the best new features in Windows 11 (and a good reason to upgrade, for those in need of one).

Step 4: Completely reinstall Windows

If none of the previous steps has fixed Windows, it’s time for a do-over. In this step, you can choose one of two different options for completely reinstalling Windows on your machine — a more draconian approach to restoring operations than any of the previous steps. Whichever method you choose, you’re going to have to reinstall all apps and applications.

Option A: Reset this PC

This repair method rolls your PC back to its built-in recovery image (usually established when the unit gets its initial Windows image and the original system/boot disk layout is established) or a fresh, clean Windows image downloaded from Microsoft’s servers. For something with the ability to change your PC so thoroughly and completely, it’s surprisingly easy to do from an administrative account.

Go to Settings > Update & Security > Recovery, and then click the Get started button under “Reset this PC,” as shown in Figure 6 for Windows 10. (You can see this same facility for Windows 11 at the bottom of Figure 5. Click Reset PC to start the process there.)

starting the reset this pc process in windows 10

Figure 6: Reset this PC rolls Windows back to its original as-delivered state.

Ed Tittel / IDG

The next screen that appears (Figure 7) tells you what’s coming. It also emphasizes that the impact of a reset operation is both major and sweeping.

reset this pc options in windows 10 keep files or remove everything

Figure 7: This is a confirmation that gives you a last chance to bail out. Otherwise, you’re in for a long process even if you elect “Keep my files.” “Remove everything” does exactly what it says.

Ed Tittel / IDG

Understand what this means: if you reset, you lose all applications and apps installed after the OS made its initial debut. You can decide whether or not to keep what Microsoft calls your “personal files” (everything in Documents, Downloads, Pictures, Music, and Videos), but everything else goes.

If you proceed with the reset, you’re presented with two options: “Cloud download” grabs fresh, new Windows installation files from Microsoft’s servers, while “Local reinstall” uses your PC’s built-in recovery image. Unless you have connectivity restraints, I recommend the cloud download option to ensure that any issues with your PC’s recovery image won’t carry through to the new installation.

For more details about using Reset this PC, see “Windows 10 recovery, revisited: The new way to perform a clean install.”

Reset this PC is convenient and requires no supplementary media, but I prefer a more traditional approach: performing a clean Windows installation from a bootable USB drive or mounted ISO, covered next.

Option B: Clean install of Windows from installation media or mounted ISO

This option means starting over with a completely new OS installation. Why might this be necessary? Aside from reasons such as incurable system instability, malware infestation, or problems that take longer to fix than the time is worth, it may be desirable to switch from legacy BIOS emulation to using available UEFI. It might also be desirable to wipe the boot/system drive clean to remove leftover items (recovery partitions, OEM partitions, and so forth) from older Windows versions and let Windows start over with a clean slate.

The good news is that as long as you’ve got a valid Windows 10 or 11 key or have activated your installation with Microsoft’s Activation servers, Microsoft will recognize your PC and permit it to activate itself without requiring you to supply a valid key. Nevertheless, if you’ve got access to the key, it’s a good idea to put that file on a USB drive or some other storage media just in case it’s requested. This seldom happens, but should it be needed, it’s good to have. (There are great tutorials at TenForums and ElevenForum about how to find and record your Windows 10 and keys, respectively.)

You can check your activation status by typing activ into the Settings search bar, then choosing “Activation settings” in the resulting list. As you can see in Figures 8 and 9, if the Activation pane says “Windows is activated with a digital license…” (Windows 10) or “Active” (Windows 11), you should be OK.

windows 10 digital license activation screen

Figure 8: If Windows 10 is activated with a digital license, you should be able to reinstall without subsequent activation problems. If they do pop up, call +1-800-426-9400 in the US to fix things.

Ed Tittel / IDG

windows 11 license activation screen

Figure 9: The Windows 11 activation is terser but no less informative. The same phone number as in Figure 8 works here, too, if you have problems to resolve.

Ed Tittel / IDG

The process for performing a clean install is 99% the same as that for the in-place upgrade from Step 3. Just boot from your installation USB flash drive, select Install Now, accept the license terms, and pick the Custom: Install Windows only (advanced) option when it’s presented to you. The rest of the details — and there are many — are all nicely covered in the TenForums and ElevenForum tutorials on this subject.

After the installation

Once you complete either the Reset this PC or the “clean install from installation media or mounted ISO” process, you’ll be starting over from scratch. For me, that means reinstalling Office plus all the apps and utilities that I customarily use on a production machine. It usually takes me 8 to 12 hours to work all the way through that process manually.

To speed the process along, I recommend using either the PatchMyPC Home Updater or Ninite utility, or using the WinGet command to import an already-exported configuration file.

Picking up the pieces after the repair

Hopefully, you’ll never have to venture further than Step 2 in this list. But if you must dig deeper than that, remember to clean up after it’s all over. That means running Disk Cleanup (or some third-party alternative) and making sure all your applications are installed and working. It also means making another backup when you’ve got things the way you want them, so you’ll have it at your disposal should things go south again.

In the meantime, enjoy your repaired and rejuvenated Windows system!

This article was originally published in June 2016 and most recently updated in July 2024.

How AI will change the way you use your browser

The modern web browser is rather staid and not terribly exciting. Google Chrome rules the roost, browser technologies like HTML and JavaScript have gone unchanged for years, and updates to browsers are more often bug fixes than new features.

But it’s about to get a little more interesting.

Generative AI is coming to every web browser. Google Chrome, Microsoft Edge, and other browsers are trying to change how we interact with the web, using AI to boil down web articles and make browsing more efficient, said Ritu Jyoti, group vice president for AI infrastructure at IDC.

“The strategy is to reinvent web search, and the hope is that you don’t have to a load a dozen browser tabs and scrutinize each one to find the information you need,” she said.

Since the dawn of the web, the way we’ve searched for content is by picking through a list of links. Now we have genAI-powered chatbots/virtual agents, which can understand plain-language questions and return direct answers. These AI assistants can also help us write and refine text, summarize web pages, translate from other languages, and more.

New chips enable (some) AI processing on the PC

All of this is made possible thanks to advances on the part of chipmakers. Both Intel and AMD have introduced a new generation of desktop, notebook, and even mobile processors with neural network coprocessors in them, as has Apple with its own custom silicon and Qualcomm, which is looking to break into the desktop/client space with its Snapdragon X line of processors.

The idea is that PCs equipped with these Neural Processing Units (NPUs) will be able to handle some genAI processing tasks, such as interactions with chatbots like ChatGPT, on the PC itself. Today, most AI workloads are cloud-based.

“They have a 40 TOPS (trillions of operations per second) floor on the NPU, which is adequate for ChatGPT inferencing. Prior efforts had well under 20 TOPS and weren’t adequate,” said Rob Enderle, principal analyst with The Enderle Group.

Intel’s current top processor (Meteor Lake generation) has around 34 TOPS across the CPU, GPU, and NPU, and its forthcoming Lunar Lake generation will top out at 50 TOPS, while AMD’s Ryzen PRO 8000 has 39 TOPS and the Qualcomm Snapdragon X Elite has 45 TOPS.

Enderle says that since these processors are coming to market just now and most AI loads are still cloud based, “this hardware anticipates movement that hasn’t occurred yet and likely won’t occur” until a critical mass of PCs with these chips are deployed.

Nevertheless, browser makers are forging ahead with AI features, some of which will take advantage of the new hardware. Most of the AI features you find in browsers now do not require the neural network coprocessor and will run just fine. But the more ambitious AI efforts will wait until there is a sufficient installed base of newer processors capable of running the features, such as large language model (LLM) processing.

AI features coming to a browser near you

Let’s look at some major and minor browsers and their AI efforts for both desktop and mobile. Some mobile browsers have not added AI as yet because the handsets don’t have the processing power, but some vendors have made it work on handsets.

Google Chrome

Among the AI features Google is rolling out in its Chrome browser is a tab organizer to create tab groups based on your open tabs. This is designed to help you manage multiple tasks simultaneously by organizing related tabs together. Chrome is also adding the ability to generate custom themes based on a subject, mood, visual style, and color of your choice. Finally, there’s the “Help me write” feature, which helps users draft content on the web, such as online reviews or queries.

These features are being gradually rolled out, with the tab organizer and AI-generated themes introduced with Chrome version M121 and “Help me write” with version M122.

Currently, these features are available in for Chrome on Macs and Windows PCs in the United States. Users need to enable these features in the “Experimental AI” page in Chrome settings.

Microsoft Edge

Copilot is Microsoft’s latest attempt at an AI assistant, replacing the far less versatile Cortana. Copilot is integrated across a huge swath of the Microsoft product lineup, including Windows, Bing, the Microsoft 365 apps, and, of course, the Chrome-based Edge browser.

Copilot in Microsoft Edge offers three key functionalities:

  1. Chat: This feature enables natural-language interactions, allowing users to engage in conversations, obtain web-page summaries, and receive answers to their queries.
  2. Compose: An AI-powered writing assistant that assists in various writing tasks, such as drafting emails or creating blog content.
  3. Insights: This tool provides contextually relevant information based on the web page you’re currently browsing, offering additional perspectives and related data.

To use Copilot, first make sure you have the latest version of Microsoft Edge installed. Click on the Copilot icon in the top-right corner of the browser window. This opens the Copilot sidebar. Type your questions in the window in the bottom of the sidebar, use one of the default prompts (such as “Generate page summary”), or click the Compose tab for writing help.

Brave

Brave is a browser built on Chrome core technology with an emphasis on privacy and blocking user tracking. The browser’s Leo AI is designed to help with various tasks such as creating summaries of web pages or videos, answering questions about content, generating new content, translating pages, analyzing information, and rewriting text. Leo supports various generative processes, such as writing letters, emails, articles, and blog posts.

Brave says chats with Leo are private, anonymous, and secure. The assistant doesn’t record conversations or use them for model training, and no account or login is required to use it.

Leo is integrated with Brave Search, so it adds AI-augmented responses into its search results, which Brave says provides more accurate and up-to-date information. Leo is built on multiple AI models, including Mixtral, Claude, and Llama.

Unlike the AI features in Chrome and Edge, Leo is also being rolled out to Android and iOS. It appears to have all the same features in the mobile apps as in the desktop version.

To access Leo on desktop, open the Brave browser, click on the three horizontal lines in the top-right corner. Select More tools > Show sidebar > Always. A sidebar will appear on the right side of the browser. Click on the Leo button in the sidebar to open the chat window.

In the mobile apps, open the Brave browser app, tap the three-dot icon and select Leo from the menu options.

Opera

When it first rolled out in mid-2023, Opera’s integrated AI assistant Aria was quite ambitious. At the time, most genAI models including OpenAI’s GPT were trained on data that was more than two years old. Aria combined GPT-based technology with current data from the web, providing more up-to-date information in its responses. However, newer genAI chatbots also use up-to-date data.

Like other AI programs, Aria assists with tasks such as explaining complex information, translating content, generating code, and answering product-related questions. Aria supports over 50 languages, is available in more than 180 countries, and works on Windows, macOS, Linux, and Android platforms.

Its Reuse and Rephrase features allow users to easily refine and build upon previous responses without retyping entire entries, and the My Style feature enables users to train Aria to write in their unique style by providing writing samples in formal, neutral, and informal tones.

To access Aria, open the Opera browser and type opera:settings in the address bar. Scroll down to “Advanced” and find the “AI Services” section. Toggle the switch to enable Aria in the sidebar.

Mozilla Firefox

Mozilla is taking a somewhat different approach to AI support in Firefox. Users will get to choose their preferred AI chatbot just like they choose their preferred search engine now. Mozilla’s initial offerings include ChatGPT, Google Gemini, HuggingChat, and Le Chat Mistral, with more to be added in the future. This feature is currently available in Nightly, an experimental version of Firefox for testing new features.

Beyond that, Mozilla is working on adding AI features that “solve tangible problems, respect your privacy, and give you real choice.” The company is looking at how it can use local, on-device AI models to enhance your browsing experience further. One feature due next quarter is AI-generated alt-text for images inserted into PDFs, which makes it more accessible to visually impaired users and people with learning disabilities.

Mozilla also says it wants to develop better alternatives to untrustworthy AI, as many chatbots have been polluted with misinformation and are being used to create deepfakes and other frauds. To that end, the company has invested $30 million in a startup called Mozilla.ai, which is focused on finding open-source, trustworthy AI opportunities and building a business around them.

Apple Safari

The Safari 18 browser will be included in macOS Sequoia and iOS 18 when they ship later this year. Safari 18 is reportedly undergoing some revamping as AI is made a part of the browser’s functionality. It’s still early, but developers testing the browser report a new “Intelligent Search” feature. This feature is believed to draw on Apple’s Ajax language-learning model to identify topics and key phrases and create page summaries.

But the bigger news from Cupertino has been around Apple Intelligence, announced at the company’s recent Worldwide Developer Conference. It’s a suite of AI tools designed to work across the entire Apple product line for tasks such as rewriting, proofreading, and summarizing text, as well as creating and editing images. Also announced at the conference: Apple’s virtual assistant, Siri, will get ChatGPT integration to enhance its functionality.

Because it is still in the early stages and months away from shipping, it’s hard to tell the role Apple Intelligence will play in the Safari browser versus the rest of the operating system. We’ll know more this fall when the new versions of macOS, iOS, and Safari are released.

One thing is clear: whichever browser you use, you’ll soon be able to get a hand from an AI assistant.

Google’s Pixel 9 and the end of Google Assistant

Let me just take a second to say what I suspect most of us are thinking right now: Sigh.

All right — with that out of the way, we’re ready to get to the crux of this conversation: Google’s next-gen Pixel 9 devices are set to officially break cover in just over a couple weeks now. And from the looks of it, they’ll be ushering in a new era of on-the-go Google goodness while, just as notably, saying adieu to an old one.

As you’ve probably figured out by this point, I’m not talkin’ about the hardware itself or anything specific to the phones’ forms. I’m talkin’ about the virtual assistant at the heart of the Pixel and broader Android experience — the thing we interact with and rely on all throughout our days and the glue that pulls all those precious Pixely pieces together.

By all counts, it looks like the Pixel 9 and its brethren could effectively mark the end of Google Assistant and put the nail in the coffin of the service Google’s spent the past eight years weaving into every last area of our lives.

One more time, together now: Sigh.

[Hey — got a Pixel? Any Pixel? Check out my free Pixel Academy e-course to uncover all sorts of useful stuff you never knew your phone could do!]

The Google Assistant era — Pixel and beyond

Time for a quick rewind — ’cause a teensy plop of context will go a long way toward paving the path we’re about to embark on.

Back in 2016, y’see, Google rolled out the red carpet for a snazzy new service that was said to be “your own personal Google” — “a Google for your world.”

That was none other than Google Assistant, of course — the service that quickly spread everywhere and started to take shape as a platform all its own.

Assistant wasn’t perfect, by any means, but it was pretty darn good at getting stuff done — helping us perform simple tasks and get swift answers without having to use our hands or futz around with any on-screen menus.

And goodness gracious, was it everywhere. That colorful Assistant logo showed up in all sorts of places — apps and physical products alike. 

As a certain greasy-beaked gazelle once put it:

It’s hard now to even convey just how big of a deal Assistant’s arrival was at that point and then continued to be over the years that followed, up until extremely recently. …

From early on, the service connected to countless other Google services — from its integration into every imaginable corner of Android and ChromeOS to its home at the core of all Google-associated smart displays, TV systems, and speakers. It came to Android Auto, even, and showed up as the branding behind all sorts of smart Pixel calling features.

Heck, Google’s presence at tech conventions turned into a literal Assistant playground for years, with endless plastering of Assistant branding everywhere you looked and — well, stuff like this:

You get the idea. (Of course you do. You’ve presumably lived through all this, right alongside me!)

Now, after eight years of expanding upon Assistant, injecting it everywhere, and turning it into a household name, Google appears poised to give up on it entirely.

It’s a facepalm-inducing pivot we’ve been we’ve watching come at us like a slow-moving train for months now, and it’s looking more and more like the Pixel 9’s arrival could be the moment it comes crashing into reality at last.

Gemini and the Google Pixel connection

For a while now, Google’s been hinting — but not outright saying — that Assistant’s days are numbered, as its newer large-language-model-powered Gemini AI assistant prepares to take its place.

That hinting has come mostly in the form of Assistant’s core functionality getting worse and worse by the week — that, and the Assistant branding Google worked to bring into every last nook and cranny slowly starting to fade while mentions of Gemini pop up everywhere. Amidst all of this, Google has said nary a word about the longer-term fate of Assistant and if or how it might fit into its Gemini-centric plans for the future.

Over the past several weeks, though, two interesting things have happened:

First, Motorola launched its latest foldable Razr phones. That in and of itself isn’t especially interesting, as no one in the right minds should be buying Moto devices these days (for reasons you, as a smart and informed Android-appreciating animal, surely know by now).

But what is interesting is the fact that those gutsy gizmos were the first Android devices to ship with Gemini present and enabled by default as the on-demand virtual assistant. No Google Assistant, in other words: When you press and hold the power button or say “Hey Google,” it’s Gemini that shows up and responds.

Second, following the typical deluge of prerelease Google hardware leaks, Google put out its own official Pixel 9 preview video this week. And — well, see for yourself:

Yup — it’s all about Gemini and its role in the Pixel 9 experience.

Combine that with the Moto device switch and everything we’ve seen about Gemini gaining prominence while Assistant sputters ever closer to the Google graveyard, and it sure seems like this could be the moment that actually marks the effective end of the Assistant era — even if it inevitably takes a while for the rest of the Android ecosystem to catch up.

Google’s Pixel 9 assistant philosophy

I won’t beat around the bush — and if you’ve read my ramblings for long, this probably won’t come as a surprise: I’m not completely convinced that the move to Gemini as the go-to Android assistant for Pixel phones or any other Android gadgets is a good thing. Not at this point, at least.

Frankly, it feels premature. Gemini still can’t handle all the basic, task-oriented stuff that Assistant excels (or at least excelled) at. And the generative-AI elements it adds into the equation — generating (awkward and robotic) text, creating (mostly creepy-looking) images, summarizing text (with varying amounts of accuracy), and so on — just aren’t all that pertinent to the Android assistant environment. That’s to say nothing of the very real questions that still linger around the quality, accuracy, and originality of the information the service serves us.

Plus, you can access all those same generative-AI elements in so many other places now — including, increasingly, within most Google Android apps — that having ’em built into your assistant seems superfluous and unnecessary. So ultimately, we’re gaining very little while losing a lot.

Now, look: I certainly get why Google wants to position Gemini as the exciting new must-have tech advancement that’s changing the face of Android and its Pixel devices. I’m just not sure the tool itself is up to that task yet. And I’m not entirely convinced it ever will be.

As our aforementioned affable antelope put it:

The reality is that large-language models like Gemini and ChatGPT are wildly impressive at a very small set of specific, limited tasks. They work wonders when it comes to unambiguous data processing, text summarizing, and other low-level, closely defined and clearly objective chores. That’s great! They’re an incredible new asset for those sorts of purposes.

But everyone in the tech industry seems to be clamoring to brush aside an extremely real asterisk to that — and that’s the fact that Gemini, ChatGPT, and other such systems simply don’t belong everywhere. They aren’t at all reliable as “creative” tools or tools intended to parse information and provide specific, factual answers. And we, as actual human users of the services associated with this stuff, don’t need this type of technology everywhere — and might even be actively harmed by having it forced into so many places where it doesn’t genuinely belong.

And yet — well, here we are.

By all indications, next month’s Pixel 9 arrival will mark the end of a major era that was positioned as the Future of Google™ a mere matter of months back. And, I suppose, only time will tell whether that ends up being a positive move from the perspective of us — the actual humans who use and rely on this technology — or a loss we’ll lament for a long time to come.

Discover all sorts of hidden magic lurking within your current Pixel phone with my free Pixel Academy e-course — seven days of advanced Pixel knowledge, straight from me to you!

Now, Spain’s competition authority is reviewing Apple’s App Store

The Comisión Nacional de los Mercados y la Competencia (CNMC), Spain’s competition authority has begun an audit of Apple’s App Store, according to Reuters.

According to the authority, Apple may have introduced different commercial conditions for developers who sell mobile applications through the App Store. If so, Apple risks fines of up to 10% of the company’s global revenue.

Apple denies any wrongdoing, saying Spanish developers of all sizes compete on equal terms on the App Store. The company says it will continue to work with the CNMC to sort things out.

The App Store has also come under scrutiny by other European regulatory agencies as well as in the UK.

What to expect with Apple’s upcoming iPhone 16

Apple introduces new iPhone models every fall. This year’s model will, of course, introduce significant support for artificial intelligence (AI) in the form of Apple Intelligence — but what else is expected from the company’s most important product? We’ve filtered through the rumor-mill to find out so you, or your business, can get some perspective on whether it’s time to update your iPhone fleet.

Things are happening in the background

In the shadows, things are moving fast. Apple Chief Operating Officer Jeff Williams visited Shenzhen, China this week. He met with senior Chinese officials and discussed various plans, including the new applied research lab the company intends to open. Call me suspicious, but I can’t help but imagine Williams might have also visited the Foxconn factory there to check the manufacturing process and sign off on iPhone production, given factories in India and in China are now recruiting staff to build the next-generation Apple smartphone. (India now accounts for 14% of global iPhone production, and that is expected to accelerate as Cupertino builds more resilience into its supply chain with its China+ strategy.)

Apple executives also seem to have visited other key companies across the supply chain in recent months, with CEO Tim Cook meeting with suppliers in China and Vietnam as recently as April. Apple announced plans to spend more on suppliers in Vietnam during that visit. While all roads don’t necessarily lead to phone, it’s not at all unlikely that component deals will be part of the conversation. That India recently relaxed some tariffs on imported smartphone components also seems relevant.

In other words, the body of (circumstantial) evidence says the Apple machine is preparing to hum into action. So, what should we expect? 

A processor fit for AI phone

The faster A18 chip is expected to be 3-nanometers, which suggests performance and battery life improvements. Power efficiency is also expected from the improved display in the devices, and it’s possible the company might seek to further differentiate its Pro family device with A18 Pro chips. 

AI is what will be driving the sales pitch here. Those A18 Pro chips will possibly feature a more powerful neural engine, making the devices the ultimate in Apple AI phones. That’s also why the RAM inside increase, up to 8GB, speculation claims.

Apple will hit the market hard, pushing home the message that these iPhones are the world’s most advanced devices for mobile AI, with a processor more capable than anyone else’s, and a private, on-device AI to help you get things done. We’ll find out the extent to which AI is being hyped in due course.

Bigger, tougher, heavier

Expect larger 6.3-in. and 6.9-in. OLED display sizes for the Pro and Pro Max models. The non-pro iPhones will remain the same size as current models, while the pros will be a little wider, taller, and a tiny bit heavier than current models. The displays might also be more scratch-resistant than before.

Apple’s spatial reality field

The 48-megapixel cameras will introduce a new vertical camera lens, likely to make the devices more capable of capturing spatial video. The iPhone 16 Pro could also gain the 5x optical zoom introduced in the Pro Mac last year.

Capture and action buttons

Speculation claims we should expect new capture (like a camera shutter) and action buttons on all the new iPhones. Apple introduced an Action button with the iPhone 15 Pro models in 2023.

Uh-oh: New prices

Driven by increasing component costs, the company is likely to include price increases across the range, so don’t be too surprised.

Battery improvements

The iPhone 16 Pro models might use stacked battery technology. This means batteries have a better lifespan and last longer between charges than before. It seems plausible Apple has shifted to this tech because it expects iPhone users to run a lot of processor intensive AI tools on the devices. There’s good news for charging, too, with 40W wired charging and 20W MagSafe charging, up from 27/15W, respectively.

Networking

The Pro models will be equipped with Wi-FI 7 and use improved versions of Qualcomm’s 5G radios, which are smaller, use less power, and support 5G Advanced. The latter is kind of important if your company is using private or sophisticated 5G services, as they are likely to work better on these devices. Standard iPhones might at last get Wi-FI 6E.

Under display Face ID?

Recent industry news in which a UK component supplier lost its Apple business hints that the company might make changes to FaceID. What those changes are remains unknown: Could these be the first iPhones to ship with under-display FaceID?

USB-C

This is inevitable. And overdue.

More to come

Expect the trickle of iPhone predictions to continue right up until Apple introduces the device in fall. Until it does, take all the speculation with a large pinch of salt. 

What about the iPhone 17?

Apple rumors are heading into prescience. We’re also being told the iPhone 17 lineup for 2025 is expected to include the iPhone 17, iPhone 17 Pro, and iPhone 17 Max. 

Apple analyst Ming-Chi Kuo says there will be a new iPhone 17 slim mode next year. He believes Apple will make a series of changes to enable that slim phone, including the introduction of a single camera lens; with a 6.6-in. display, A19 chip, and Dynamic Island, the device should still impress. What we don’t know is whether the 5G chip Apple plans will be one made by Apple. Kuo seems to think it will — and that this will be the most expensive iPhone you can buy.

Please follow me on Mastodon, or join me in the AppleHolic’s bar & grill and Apple Discussions groups on MeWe.