Archive for May 21, 2024

Tuesday, May 21, 2024

Windows Copilot+ AI Features

Microsoft (Hacker News, MacRumors, Ryan Jones):

Now with Recall, you can access virtually what you have seen or done on your PC in a way that feels like having photographic memory. Copilot+ PCs organize information like we do – based on relationships and associations unique to each of our individual experiences. This helps you remember things you may have forgotten so you can find what you’re looking for quickly and intuitively by simply using the cues you remember.


Combine your ink strokes with text prompts to generate new images in nearly real time with Cocreator. As you iterate, so does the artwork, helping you more easily refine, edit and evolve your ideas. Powerful diffusion-based algorithms optimize for the highest quality output over minimum steps to make it feel like you are creating alongside AI. Use the creativity slider to choose from a range of artwork from more literal to more expressive.


Live Captions now has live translations and will turn any audio that passes through your PC into a single, English-language caption experience, in real time on your screen across all your apps consistently. You can translate any live or pre-recorded audio in any app or video platform from over 40 languages into English subtitles instantly, automatically and even while you’re offline.


Eye contact teleprompter helps you maintain eye contact while reading your screen. New improvements to voice focus and portrait blur help ensure you’re always in focus.


Every Copilot+ PC comes with your personal powerful AI agent that is just a single tap away on keyboards with the new Copilot key. Copilot will now have the full application experience customers have been asking for in a streamlined, simple yet powerful and personal design. Copilot puts the most advanced AI models at your fingertips. In the coming weeks, get access to the latest models including GPT-4o from our partners at OpenAI, so you can have voice conversations that feel more natural.

Dare Obasanjo:

Recall reminds me of Stuff I’ve Seen, a 2003 Microsoft Research project to help solve the problem of finding content you’d previously seen. The big problem then was most stuff you saw was on websites not local files.

Recall uses screenshots to solve this.

Matt Birchler:

If you saw this feature and thought, “huh, that sure looks like Limitless,” you would be absolutely right. Just a few weeks ago I suggested Apple should buy Limitless and build it into macOS natively, but Microsoft beat them to the punch by just building it themselves.

It’s an absolute classic Sherlocking, but it totally makes sense. The second I saw Rewind 2 years ago I knew it was something cool, but that was exactly the sort of feature that only works for more people if it’s built by the OS provider. Microsoft is already dealing with privacy concerns with it, so you can only imagine how people feel about letting a VC-funded company they’ve never heard of record everything they do, even if it’s all local, all encrypted, and theoretically actually private in the way people want.

John Gruber:

Recall can “view” and remember everything that appears on screen because it’s integrated with the Windows 11 graphics system. That’s the sort of “AI feature” that truly benefits from being a first-party solution that can integrate at lower levels of the OS than third-party apps can.

Rui Carmo:

I’m a bit skeptical on the concept (even though I did use Windows 10 timeline a fair bit), but I find it rather telling that a key future Windows feature is tied to ARM processors (plus their NPUs, sure, but it’s a key sign that Intel lost the plot here).

Ben Thompson:

That celebration, though, is not because Windows is differentiating the rest of Microsoft, but because the rest of Microsoft is now differentiating Windows. Nadella’s focus on AI and the company’s massive investments in compute are the real drivers of the business, and, going forward, are real potential drivers of Windows.


Nadella, similarly, needed to break up Windows and end Ballmer’s dreams of vertical domination so that the company could build a horizontal services business that, a few years later, could actually make Windows into a differentiated operating system that might, for the first time in years, actually drive new customer acquisition.


Update (2024-05-29): Nick Heer:

Recall is the kind of feature I have always wanted but I am not sure I would ever enable. Setting aside Microsoft’s recent high-profile security problems, it seems like there is a new risk in keeping track of everything you see on your computer — bank accounts, a list of passwords, messages, work documents and other things sent by a third-party which they expect to be confidential, credit card information — for a rolling three month window.

See also: Bruce Schneier and Ben Thompson.

Microsoft’s Copilot+ PCs

Tom Warren (MacRumors, Hacker News):

Over the past two years, Microsoft has worked in secret with all of its top laptop partners to ready a selection of Arm-powered Windows machines that will hit the market this summer. Known as Copilot Plus PCs, they’re meant to kick-start a generation of powerful, battery-efficient Windows laptops and lay the groundwork for an AI-powered future.

“You’re going to have the most powerful PC ever,” says Yusuf Mehdi, executive vice president and consumer chief marketing officer at Microsoft, during the briefing. “In fact, it’s going to outperform any device out there, including a MacBook Air with an M3 processor, by over 50 percent on sustained performance.”


One of the big advancements is an improved emulator called Prism, which Microsoft claims is as efficient as Apple’s Rosetta 2 translation layer and can emulate apps twice as fast as the previous generation of Windows on Arm devices.


Overall, Microsoft believes 87 percent of total app minutes spent on these Copilot Plus PCs will be inside native apps.

They also claim to have significantly better battery life.

Martin Pilkington:

The Snapdragon X Elite benchmarks are impressive, but when you realise it’s using 80W to slightly beat the M3 Pro which is using under 50W for the same benchmark (and both are matched by the M4 which is probably using much less than the Pro) I don’t think Apple it too worried.

Andrew Cunningham:

The Surface Laptop—referred to as the “7th edition” in its Microsoft Store URL but simply called the “Surface Laptop” most other places—is Microsoft’s first traditional laptop with an Arm chip. The laptop comes in both 13.8-inch and 15-inch sizes and starts at $1,000 for a 13.8-inch config with a Snapdragon X Plus chip, 16GB of RAM, and 256GB of storage. The cheapest 15-inch version is $1,300, but it includes a Snapdragon X Elite chip instead.


As for the Surface Pro tablet, this update to Microsoft’s flagship convertible is a lot closer to what Microsoft shipped a year and a half ago in the Surface Pro 9 and Surface Pro 9 with 5G. The new Surface Pro, called “11th edition” in its Microsoft Store URL but not in most other places, still weighs just a hair under 2 lbs, still has the same dimensions (and maintains compatibility with the same Slim Pen and keyboard covers), and still has a 13-inch screen. It starts at $1,000 for a version with a Snapdragon X Plus chip, 16GB of RAM, 256GB of storage, and an IPS LCD display; keyboards and pens are still add-on accessories.

Martin Pilkington:

Looking at the new Copilot + PC specs I hope they push Apple to move to a minimum of 16GB of RAM on M4 Macs (especially given how much more powerful the GPU and NPU are)

John Gruber:

Are any of today’s first batch of “Copilot+ PCs” fanless? If not, can any of them truly be said to have “taken aim” at the MacBook Air?

John Gruber:

I’ll go out on a limb and say that today marks the beginning of the end for x86. Either the x86 architecture has reached an inevitable endpoint, or Intel and AMD are just unable to compete talent-wise. (Or both.) But as of today the performance-per-watt gulf between ARM and Intel/x86 is no longer just an Apple silicon thing — it’s now a PC thing too.


The saddest part of the event were the cursory appearances — both by pre-recorded videos, despite it being an in-person event in Redmond — of Intel CEO Pat Gelsinger and AMD CEO Lisa Su. Their token appearances felt like Microsoft pretending they haven’t moved on from x86, during an event whose entire theme was, effectively, “moving on from x86”.


Apple Updates Silently Enable iCloud Keychain

Jeff Johnson:

I’ve discovered today that unfortunately this issue—this bug, I would call it, though who knows whether Apple considers it a bug or “expected behavior”—still exists with the latest versions of macOS Ventura and Sonoma, 13.6.7 and 14.5 respectively.


The external drive had a macOS Ventura 13.6.7 boot volume with iCloud enabled but iCloud Keychain disabled. After updating the volume to macOS Sonoma 14.5, iCloud Keychain was enabled. (I then disabled iCloud Keychain, which actually caused System Settings to hang and eventually crash, but afterward iCloud Keychain did seem to be disabled.)


What I’d like to do is update from Ventura to Sonoma without an internet connection, giving Sonoma no chance to upload my passwords or other data to iCloud before I can disable iCloud Keychain.


You might wonder why I don’t sign out of iCloud before I update from Ventura to Sonoma. It turns out that there’s no point in that, due to another bug, “Signing out of iCloud and signing back in again forgets all of your previous iCloud settings” (FB12168173), which I also discovered last year.

Because installing macOS also re-enables Wi-Fi, his workaround was to turn off Wi-Fi after downloading the installer, delete his Wi-Fi password, and then install the update.


If you’ve never enabled iCloud Keychain and recently upgraded to iOS 17, chances are good that your passwords are now stored on Apple servers. As confirmed by many users, iOS 17 secretly turns iCloud Keychain on. This video shows the entire process step by step[…]


Update (2024-05-28): See also: Hacker News.

Update (2024-05-29): Marcin Krzyzanowski:

I noticed my disk storage went drastically low and I started to check system, then I realized something ( #macos update???) enabled iCloud Photos synchronization to my Mac (that can take all the storage it get, and for that very reason I didn’t enable it on my mac)

Update (2024-05-31): See also: TidBITS-Talk.

Update (2024-06-03): Johann Campbell:

Really wish Apple could stop toggling iCloud Photos on without my permission, when it KNOWS I won’t pay for more than the base 5 GB of iCloud storage.

Update (2024-06-05): Jeff Johnson:

A follower on Mastodon gave me a nice tip on how to prevent this in the future: create a configuration profile.

First, download the Apple Configurator app from the Mac App Store. Then open Apple Configurator, select New Profile from the File menu, uncheck Allow iCloud Keychain in Restrictions, and save the .mobileconfig file.

Slack AI Privacy

Ashley Belanger (Hacker News):

After launching Slack AI in February, Slack appears to be digging its heels in, defending its vague policy that by default sucks up customers’ data—including messages, content, and files—to train Slack’s global AI models.

According to Slack engineer Aaron Maurer, Slack has explained in a blog that the Salesforce-owned chat service does not train its large language models (LLMs) on customer data. But Slack’s policy may need updating “to explain more carefully how these privacy principles play with Slack AI,” Maurer wrote on Threads, partly because the policy “was originally written about the search/recommendation work we’ve been doing for years prior to Slack AI.”

Maurer was responding to a Threads post from engineer and writer Gergely Orosz, who called for companies to opt out of data sharing until the policy is clarified, not by a blog, but in the actual policy language.

Gergely Orosz:

An ML engineer at Slack says they don’t use messages to train LLM models. My response is that the current terms allow them to do so. I’ll believe this is the policy when it’s in the policy.

Richard Speed:

Salesforce division Slack has responded to criticism by users outraged that its privacy principles allowed the messaging service to slurp customer data for AI training unless specifically told not to, claiming the data never leaves the platform and isn’t used to train “third party” models.

The app maker said its ML models were “platform level” for things like channel and emoji recommendations and search results, and it has now updated the principles “to better explain the relationship between customer data and generative AI in Slack.”


The privacy principles were overhauled in 2023 and contained the text: “To develop AI/ML models, our systems analyze Customer Data (e.g. messages, content and files) submitted to Slack.”


The principles have since been tweaked slightly, and now read: “To develop non-generative AI/ML models for features such as emoji and channel recommendations, our systems analyze Customer Data.”

Adam Engst:

If people actually read Slack’s privacy principles document instead of just reacting to an incorrectly titled link or an out-of-context screenshot on X/Twitter, they would see that Slack isn’t doing any of those things.

However, the “unambiguous sentences” that he quotes seem to be from the current privacy principles, not the May 17 version that sparked the outrage.

More seriously, there’s an important point to make here. Even as we rely ever more on gadgets and services, society has lost a great deal of trust in the tech industry. This controversy arose because the suggestion that Slack was doing something underhanded fit a lot of preconceived notions.

People didn’t want to give them the benefit of the doubt because their behavior played into preconceived notions and seemed sketchy. Their privacy document was antiquated (written to cover a previous AI feature) and not very clearly written. It gave examples of how the customer data might be used but didn’t specify limits. The document has no modification date or change history, with the overall privacy policy still showing a date of July 5, 2023. You had to opt out, and not via a visible setting—but by sending them an e-mail with a special subject. It’s all basically the opposite of what Steve Jobs recommended.

Update (2024-05-22): Adam Engst:

All that said, I still feel like Slack’s mistake in failing to update the document to be more clear wasn’t that bad. The subsequent changes Slack made show that even if the document wasn’t as clear as would be ideal, Slack wasn’t trying to put one over on us. Even in the problematic May 17 version, Slack said:

For any model that will be used broadly across all of our customers, we do not build or train these models in such a way that they could learn, memorise, or be able to reproduce some part of Customer Data.

Of course, because of the lack of trust many people have in the tech industry, even relatively clear statements like that don’t necessarily have the desired effect. “Sure,” one may think, “that’s what you say, but how do we know that’s true?”

And we don’t. There are many lapses, security breaches, and broken promises. But simultaneously, we have to trust the technology we use to a large extent because the only other option is to stop using it.