It’s not straight copying, no. It’s lossy compressing (think JPEG) and then decompressing. Is a JPEG a copy, even though it’s not 100% accurate? If i blend with millions of other lossy images, I don’t have an actual copy of any of them. What do I have instead? A representation of the entire catalog/library?
Siri 2.0 - Apple and generative AI
- Thread starter wrylachlan
- Start date
OrangeCream
Ars Legatus Legionis
Um, yes, a jpeg is a copy. Copyright infringement is an old subject now. Do you think it was only applied in situations where perfect copies exist?
Via Mac Rumors, there’s a bunch of new Gurman rumors about the AI stuff Apple will be revealing at WWDC next week:
- It’ll be called “Apple Intelligence.” (Groooooan.)
- It’ll work on Macs and iPads with an M1 or better (yay!), but only iPhones 15 Pro and 16 (boo!)
- The features will be opt-in (Giant YAY!)
- The AI stuff will gobble up about half of the WWDC keynote on Monday.
Color me genuinely surprised that they’re keeping M1 machines in the mix for this long.
Limiting it to 15 Pro and up feels like a regression to the first 5-ish years of iPhones. People have been kind of spoiled by pushing the iPhone announcement to September. When new phones were announced at WWDC, it felt like every generation came with exclusive software features, but after the events were split, it just became cameras and speed.
I can't imagine Apple would leave out the vast majority of their users. I'm sure older iPhones will be supported, just in a limited capacity
The rumors have been stable that it will be an iPhone 16-series tentpole.
I was shot down aggressively in another thread for suggesting that Apple arbitrarily gatekeeps features as a marketing construct. Viewed through that lens, I am surprised to see the M1 receive support (assuming Gurman has the goods), but limiting iPhones seems par for the course. We'll see!
Curious about iPads, though. Would this suggest any iPad with an M1 or newer will be supported?
EDIT: Not sure where I read this, so grain of salt and all that, but I saw a theory that 8GB of RAM is the dividing line, not just a modern Apple-derived processor with a neural engine. So that would suggest any Apple Silicon Mac, the iPhone 15 Pro and up (but not the iPhone 15 or anything below), and any Mx-based iPad.
The willingness to extent support to any iPad and Mac with an M1 suggests that this might not be an arbitrary gatekeeping marketing construct, but rather actually based on hardware limitations.
According to the article, yes: “ For those using iPads or Macs, models will need an M1 chip at least. “
ah, thanks! I've exhausted what I can read with my free Bloomberg account so I wasn't able to see past the lede and didn't want to assume MacRumors' summary was 100 percent accurate.
Oh, sorry—I’m just going by MacRumors too; I can’t get to the underlying Bloomberg article either.
FWIW, I can’t remember ever being led astray by MacRumors’ summaries before. (Rumors of course haven’t panned out, but I don’t recall them ever misrepresenting the rumors themselves.)
Always is. Comes down to "doesn't run well enough", and while you can quibble over what "enough" means it's never a "because we're jerks" answer.
It’s so obvious now that I see it, I wish I had thought of it. It’s still “AI,” but it’s not just Artificial Intelligence like everyone else is doing, it’s Apple Intelligence that only Apple devices can do. It’s a fantastic marketing name.
If this is true, I’m surprised that an M1 chip from 2020 will be able to handle AI tasks that an A16 Bionic chip from 2023 can’t. Maybe it has more to do with the amount of RAM? Only the iPhone 15 Pro (and presumably iPhone 16) has 8 GB of RAM, whereas all Macs have 8 GB minimum.
They need to careful with that as it can backfire ....
The obvious question will likely be - So, is Siri a great example of Apple Intelligence?
The obvious question will likely be - So, is Siri a great example of Apple Intelligence?
Consensus is that Apple Intelligence requires the 8GB of RAM included in its support SoCs.
Standing outside the ”AI” umbrella we have a handful machine learning features, including web page highlights and summaries, mail categorization, voice transcription in notes, and the fix-my-handwriting Smart Script feature on iPad, one I’m personally very interested in.
Alas I was disappointed to learn this feature will not be supported on my 4th-gen (A12Z-based) iPad Pro. Curiously, though, the feature is supported on several non-Pro A14-based iPads.
Why might this be? The A14 is a 4GB SoC, while all A12Zs include 6GB of RAM. It so happens that the A14 introduced the 16-core neural engine; all ANEs before it featured only eight cores. <sigh>
Anyway, just pointing this out as the first ML feature where evidence points to the ANE itself, not RAM, as the constraining factor.
Standing outside the ”AI” umbrella we have a handful machine learning features, including web page highlights and summaries, mail categorization, voice transcription in notes, and the fix-my-handwriting Smart Script feature on iPad, one I’m personally very interested in.
Alas I was disappointed to learn this feature will not be supported on my 4th-gen (A12Z-based) iPad Pro. Curiously, though, the feature is supported on several non-Pro A14-based iPads.
Why might this be? The A14 is a 4GB SoC, while all A12Zs include 6GB of RAM. It so happens that the A14 introduced the 16-core neural engine; all ANEs before it featured only eight cores. <sigh>
Anyway, just pointing this out as the first ML feature where evidence points to the ANE itself, not RAM, as the constraining factor.
Last edited: