Apple Intelligence doesn't use OpenAI at all. Siri and writing tools can tap into ChatGPT to compose text, e.g. create a story but that requires approval from the user. Claiming that all Apple did was "hastily shoved together selection of OpenAI API calls" is a misrepresentation of what Apple showed.
I think you disagree with yourself? How is Apple AI not using ChatGPT if Siri uses ChatGPT? You said it yourself. Heck, Apple said it themselves.
The fact that there is a prompt asking for permission doesn't change that. Maybe some parts of the AI in iPhone don't use ChatGPT but others do, as advertised.
> Apple Intelligence doesn't use OpenAI at all. Siri and writing tools can tap into ChatGPT to compose text, e.g. create a story but that requires approval from the user.
Apple Intelligence is run on-device and on Apple's Private Cloud Compute, it does not use OpenAI. Apple also allows ChatGPT (and later other models) to be called directly from Siri and writing tools, in cases where Apple Intellgence can't solve the task. That is for example for when some text should be composed, like a story or recipe. Nothing about this contradicts with what I have written previously.
Yeah so it seems we agree, Apple resorts to third party AI for some tasks.
"Apple Intelligence" is a strange marketing driven name because it implies to the tech naive that all AI in an iPhone is offline.
When in fact "Apple Intelligence" is just the name for some of the AI functionalities in Apple devices. Some functionalities will still send data to ChatGPT and perhaps other third party providers in the future.
But Apple gets to advertise "Apple Intelligence" as privacy conscious, offline, personal AI and still be technically correct. Classic.
It's very clear when ChatGPT gets involved - you get a prompt asking you if you want to send the query to ChatGPT ("ChatGPT" being referred to by name). There's no feasible way to have your data sent to ChatGPT unintentionally.
Anyway, ChatGPT was a very small portion of the Apple Intelligence demo. The significant majority of the functionality they showed (all the non-ChatGPT stuff) was on-device or in Apple's Private Cloud Compute, the privacy story for which is quite good.
I did read the article. The person I'm replying to claims the entire debate was "uninformed hysteria", which means they thought the previous security model already required admin.
And in a way same applies to M3 vs M4 Geekbench scores. A few new instructions were added. Aside from those it's nowhere near the 25% improvement there either.
Without SME, M4 still scores in the 3800s which is still significantly higher than Zen5's projection. By the way, M4 is in a 5.1mm fanless tablet.
Zen3 to Zen4 had an even larger increase in Object Detection score in GB6. From Zen3 to Zen4, Object Detection increased by 2.5x due to AVX512 which is more than M3 to M4's 2x increase.
Eventually they can just re-enable it as part of an "important update". They can even stop you from being able to disable it entirely. If you're using Windows, it's not really your computer. At least you aren't the administrator of that device. Microsoft can access any file, install any software, change any setting, or remove any access at any time for any reason with no notice or indication to you that it happened. They can even shutdown your computer. Any device which works like that is not one that's under your control.
Every internet connected windows computer is insecure by design and cannot be trusted to protect your privacy or security.
> Microsoft can access any file, install any software, change any setting, or remove any access at any time for any reason with no notice or indication to you that it happened. They can even just shutdown your device. Any device which works like that is not one under your control.
Having this power is inherent in making the OS. Whoever is the vendor of your particular Linux distribution has the same powers, it is just that you trust them not to use them (or, in a very small theoretical minority of cases, you’ve audited the code and binaries yourself).
So yes, you shouldn’t use an OS from a vendor you don’t trust, I agree completely.
I don’t understand why people are acting like this is earth shattering news though, this has always been the case since people started using software they didn’t write themselves.
No, it really isn't. For decades I owned computers with operating systems which didn't have that capability. Once installed and configured, the OS was consistent and (reasonably) stable. Someone would literally have to break into my house or office to modify my settings or install software against my wishes.
Even after I started connecting my devices to the internet the OS itself had no ability to do these things and couldn't gain that ability unless I explicitly chose to install updates that enabled that behavior. That's entirely different from the situation today where MS forces updates and restarts, installs unwanted software on our computers, and has files and folders that we (even using administrator accounts) don't have access to.
Linux too is very different. Linux is transparent about what it does, adds, or changes. You have the power to choose which updates to apply or not. You have the power to modify any part of your OS so that it does what you want. I can't speak to all distros out there, but I've never seen a linux system force a restart in the middle of the day, or reinstall applications users removed without notice. Can't say the same for Windows. Unlike Windows, linux typically respects its users and their wishes.
You really don't have to write your own software in order to have software that respects you and leaves you in control of your own devices. It's kind of crazy that you'd think there could be no other way.
> My point is that there isn’t a technical reason that prevents Linus distros, or any other OS, from restarting your computer whenever it feels like it.
Go install MS-DOS 6.22 on a computer. You can leave that system up and wait your whole life and you'll never see it suddenly restart your computer without asking. The technical reason why it can't is because there is no code in that OS designed to check for and accept an order from someone at Microsoft to restart your machine without asking. It doesn't exist. You could choose to find or write and then install new software that gives that OS the capability to do it, but that capability just isn't there otherwise.
There's no rule that an OS has to include code to violate the rights and will of the people who install it on their devices. That's a choice that MS made. Far too many people have accepted that behavior from them so they keep pushing and pushing with new and increasingly user-hostile code and behavior but none of that is inevitable or unavoidable. That is what's a very recent development. For a very very long time no operating system would have dared to violate their users that way. None of them did.
Yes, at a certain level you have to be able to place some level your trust in your OS. Especially one with internet access. MS has shown themselves to be entirely untrustworthy, but they could still change all of that. They could strip out every line of code that allows them to remotely access your system without your explicit permission. They could be 100% transparent about what their updates will do to your computer if they are installed and they could give you the ability to not install any update you didn't like and revert to any previous state. They could give you full access to every file and directory and process and give you the ability to control every aspect of their OS. They could vow to never modify a setting after you've changed it. They just choose not to do those things, because they don't care about you or your privacy or your wishes, or your rights. As long as people continue to use windows, Microsoft stands to make a lot of money by ignoring those things.
Right. There is no technical reason why the OS vendor couldn’t attack you in the past, but software industry norms have changed over the years. What has changed is trust.
Today, you have to consider commercial OS vendors (and third party application developers) to be remote attackers in your threat model. More and more, they write their software to serve themselves rather than their users, and to make computers do what they want them to do, not what the users want them to do. This was not the case decades ago, even if the technical ability was there all along.
> More and more, they write their software to serve themselves rather than their users
Well said! I really miss when our products served us but I can't think of a recent purchase of anything internet capable that wasn't designed to work for someone else (and against me no less). I don't see "never own an internet capable product again" as a viable option here, and I'm not sure what else we can do to protest this besides push for government intervention. In the meantime, I try to firewall off whatever I can.
> My point is that there isn’t a technical reason that prevents Linus distros, or any other OS, from restarting your computer whenever it feels like it.
Wrong, the point of the operating system is to manage local state, hardware, etc.
The point of viruses, malware, and spyware is to exfiltrate data and control from a set of systems. This is getting to the point where Windows itself is a worse virus than just downloading the random shady program from the internet, with all anti-virus turned off...
And the technical distinction? You can turn off everything in linux, you can make it so the computer cannot update itself. The Operating System is unable to change itself in this configuration, the only way around this is for you to choose to update it.
This cannot be done with Windows, not without resorting to technical tricks that look at lot like what malware and viruses have to do. This a is pretty, and important technical distinction:
Operating Systems don't have built-in backdoors that you cannot turn off by design.
Malware and botnets, have built-in backdoors that you cannot turn off by design.
> Wrong, the point of the operating system is to manage local state, hardware, etc.
Yes, and to manage local state and hardware it needs to be able to control the hardware and other software.
You can build an OS that doesn’t take advantage of those capabilities but you can’t build an OS that doesn’t have them. Hence why the key is trusting your OS vendor.
> And the technical distinction? You can turn off everything in linux, you can make it so the computer cannot update itself. The Operating System is unable to change itself in this configuration, the only way around this is for you to choose to update it.
Sure you can do all that but what you can’t do is make it so your Linux based OS can’t control your hardware and software. At the end of the day, the key is still trust, either in your vendor or in your own audit.
You have presented a great many reasons why Linux is more trustworthy than Windows to many people but you cannot get around the problem of having to trust someone.
> but you cannot get around the problem of having to trust someone.
You still don't get it...
At the end of the day, I don't have to trust anyone with an OS that I fully control, with hardware that I fully control, because I can verify every bit of hardware, every bit of software, even stop the kernel from doing things if I want to (yes its possible, technically).
Sure, I can place some temporary trust in some components, but it doesn't matter really, because I can always swap/disable/remove audit/reaudit any component. You can choose to trust, as much or as little as you want. I don't have to use the kernel at all if I don't want to, I could swap in another one and still be good to go (more or less).
This is different from the case here, where by default, not of my choosing, actively and persistently nearly every aspect of a Windows computer is obfuscated, un-auditable, actively and without consent doing things that are not operating system things but spyware, bloatware, crapware, or just straight up malware. You can wave your hands around as much as you like waffling about "trusting someone" but there is a big big difference between someone acting reasonably, and choosing to allow them into your home, and "trusting" someone with a knife to your back not to shiv you.
One is reasonable, a choice, and low risk, the other is clearly none of those things. You don't have to "trust" low risk situations, they are just low risk, no trust involved.
There's still a big difference between "a surreptitious hack is technically possible with future development and getting you to accept a bad patch" versus "the company is actively using sketchy powers and trying to make them constant and socially normalized."
In one case, someone discovering sketchy secret backdoor code causes a huge flap and damage to the company's brand and stock price etc.
In the other, some corporate drone bafflegabs about it enabling superior customer satisfaction synergies, while pointing to a tiny clause in an enormous contract of adhesion to claim everybody knowingly agreed to it.
Yes, if you compile your own binaries, audit the source code, and for good measure audit your compiler and the system you are using to compile it, then there is a meaningful difference*.
Since 99% of users don’t actually do any of that, then in practice there isn’t actually a difference.
* I am aware that there are shades of grey between the scenario I describe and proprietary software - I am just being hyperbolic for rhetorical reasons.
> Since 99% of users don’t actually do any of that, then in practice there isn’t actually a difference.
I understand the hyperbole, but in practice we have strong evidence that MS is willing to intentionally use their OS against you, while we don't for your typical linux OS. That really means a lot.
When linux distros disrespect their users even a little (see for example https://www.pcworld.com/article/436097/ubuntus-unity-8-deskt...) users really don't put up with it and they can switch to another distro with very very little effort/change and even have the ability to modify the source and fork the OS. That helps to keep people a little more honest.
The backdoored compiler problem is a bit harder. We can write our own, but it's turtles all the way down. Increasingly we also have to put a lot of trust in our hardware. There are only a small number of companies making CPUs and wireless chips. I imagine they're under enormous pressure from governments to compromise the privacy and security of the people using that hardware and we have less trust in our own devices the more we have "trusted computing" forced on us.
I know how you feel. I was a fan of DOS, Win89SE, Windows 2000 Pro, and Windows 7 Pro (until 7's updates started including Win10's invasive telemetry). The good news is that alternatives are better than ever and the few windows applications I still use can run using wine (or worst case a VM)
Yeah this is my problem with windows. I’ll delete or disable things that were added to my machine only to have windows update restart my computer and those things show up again. I’m using a legit copy of Windows 11 Pro and it’s absurd that I’ve had to delete or disable random shit like social media apps multiple times.
For some reason vim-airline in combination with neovim 0.10 has really bad performance when scrolling large files. This doesn't happen with neovim 0.9 and regular vim.
Lockdown mode is used to protect journalists or other people against malware like Pegasus. It doesn't get activated by being in an unusual location, it has to be manually activated in settings.
Heya! Sorry, I would've added some links but I was on mobile and still am.
As I hinted at, this is mostly coming from anecdotal wisdom I've heard. The hardest evidence seems to be Apple's chips increasing more (10-20%) in GB6 vs GB5 when compared to how much other manufacturers such as QLC and INTL increased. As another comment said, GB5 is well known to be quite accurate with SPEC (industry standard), thus this discrepancy with GB6 and GB5 is a bit concerning.
Of course synthetics are always skewed (sometimes because a chipmaker tries, sometimes not), most famously Cinebench which has lost most credubility for favoring SIMD perf far too much.
Also might be worth it to check out the top comment to see an example of how a choice can bias syncthetics but not real-world.
I would recommend reading the Geekbench 6 internals document, they explain the rational behind the change.
> Geekbench 6 uses a “shared task” model for multi-threading, rather than the “separate task”
model used in earlier versions of Geekbench. The “shared task” approach better models how
most applications use multiple cores.
> The "separate task" approach used in Geekbench 5 parallelizes workloads by treating each
thread as separate. Each thread processes a separate independent task. This approach scales
well as there is very little thread-to-thread communication, and the available work scales with
the number of threads. For example, a four-core system will have four copies, while a 64-core
system will have 64 copies.
> The "shared task" approach parallelizes workloads by having each thread processes part of a
larger shared task. Given the increased inter-thread communication required to coordinate the
work between threads, this approach may not scale as well as the "separate task" approach.
Nothing about this is biased towards Apple. GB6 simply scales worse with more cores due to increased inter-core communication requirements.
This is correct. AMD and Intel CPUs had many slower cores. GB5 made them look better. Apple has fewer cores but more fast ones.
Most applications can't utilize many cores. Thus, usually, consumer applications perform better with fewer but faster cores than many slow cores. Geekbench is a consumer CPU benchmark.
I think you may be a few steps behind, what you're explaining is the rationale for the ST and MT individual scores. This is something all modern benchmark software has.
Hallo! That was really interesting to read through!
I just wanted to clarify that by bias I mean a skew towards one side that causes scores to misalign with SPEC. In this case, this bias against high-latency ICC is one of the causes of such 'bias'.
Thanks for the thoughtful and informative response though.
