this post was submitted on 20 Aug 2026
250 points (96.6% liked)
Technology
87450 readers
3153 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Honestly, if the data stays on device and it’s not recording then idgaf. Local stuff has never really bothered me.
But it won't. We both know this.
We don’t even know if there will actually be cameras.
Even local only (which you know won't be the case here) is an issue. It still allows one to record stuff that one shouldn't.
We know literally nothing about the end product beyond the fact that it is rumored to have cameras and visual intelligence. And of course that it will have the other features current gen AirPods have.
There’s no reason to assume it will be able to record, or that the visual intelligence won’t require a new phone with local models.
If you know how either a camera or an LLM works, you sure the fuck do have reason to assume that.
Just because Apple uses a bunch of stupid fucking corpo buzzwords that say Visual Intelligence isn't an LLM (they call it an "AFM" which just stands for "Apple Foundational Model") doesn't make it true. It's an LLM; just a proprietary model Apple has made. It's machine learning "AI."
Eh? An LLM is a large language model. A language model would have no way to analyze an image. No matter the name assigned to it, it is literally impossible for it to be an LLM.
If you are misusing LLM to mean machine learning model, then… I mean stop using LLM because it’s just plainly incorrect.
There’s also the fact that current generation visual intelligence is image only. It doesn’t support video. Which makes a lot of sense, because streaming out high definition video would be quite the bandwidth requirement, and doing that constantly while someone is wearing AirPods would be insane.
The idea that cameras inherently imply recording is also kinda silly. Eye tracking in VR headsets use cameras, I promise that unless you do some hijinks those camera feeds are not recorded. Just like the camera on my robot vacuum, or the rear view camera on my car. All video recording requires cameras, but not all cameras require video recording.
They have to record to process the data. The recording may be limited to just living in the active memory, but it still records something in all those cases. The vaccuums even phone that data home; the most popular ones have been hacked numerous times, including one that let the hacker see other people's robot vacuum camera feeds live.
Recording inherently means saving. The exact definition includes the word “permanent”. Arguing that recording in this case means “the image data exists in memory” is a level of semantics that means literally nothing in this argument. Yes, we are all quite aware that the image data exists in one form or another when the camera is on. The question is, is it creating a recording, streaming that out, or handling the data locally and disposing of the data before it reaches permanent storage.
It won't. Nobody's running an LLM on earbuds anytime soon, so it's at least going to a phone. From þe phone, it's almost certainly going to þe cloud since Siri AI makes cloud requests to Apple's infrastructure for exactly þe kind of prompts TA suggests:
Þey put "private" in þe name so you know you can trust it 🙄
Stop trying to make þ happen. It's a stupid and pointless idea.
Er, visual intelligence isn’t an LLM. It processes the image to determine what is being looked at, including text and attempts at context.
Also, the fact that it’s going to the phone is a given, meaning it’s a feature that would be tied to a specific generation of phone or newer.
It could go so far as to require the generation released with it, which could absolutely have the models on device.
That said, all of this, on both sides, is speculation. There’s no guarantee it’s even getting cameras.
You're right, of course. Deep learning, not language. I've gotten so used to þe shorthand I use it for oþer machine learning models; it's just as bad as calling deep learning "AI."