Apple barely talked about AI at its big iPhone 17 event

I think this is the correct approach on a phone. I don't want AI front-and-center. I want it in the background quietly making everything better. To me, that's a much more useful form of AI.

I have a similar philosophy about home automation. Every few years I geek out and set up a bunch of crap and spend a bunch of money and waste a lot of time, and then it tends to fall apart fairly quickly and I repent of everything...

...except for the motion-activated lighting in our foyer and laundry room. $15, 15 minutes to install, no additional charges, no external services, no security issues, and just works year after year with 100% reliability.

"Any sufficiently advanced home automation is indistinguishable from a haunting."

And I want it turned the fuck off, quietly not doing anything with my personal shit.

I want to reach for my tools when I want to use them.

I'll argue that face recognition, event detection and share recommendations are nice features.

They are all done locally on your device for the last decade, at least.

I think it is just generally the correct approach. It was not that long ago that ML (outside of research) was often talked about but it being ML was not the focus. The focus was on the actual benefits and what it did.

Of course this is going to be spun and turned into a negative, but I basically want ML to be invisible again. The benefits being clear, but the underlying tech no longer mattering.

[flagged]

that stuff is also essentially machine learning, just more parameters and better marketing

You're letting the hype men set the goalposts for you, then, as every ML thing has been retroactively rebranded as "AI."

Remember the term "smart" as applied to any device or software mode that made ~any assumptions beyond "stay on while trigger is held"? "AI" is the new "smart." Even expert systems, decision trees, and fulltext search are "AI" now.

> You're letting the hype men set the goalposts for you, then

Not really, I'm taking the hint. If they call a feature "AI", there's a 99% chance it's empty hype. If they call a feature "machine learning", there may be something useful in there.

Notice how Apple, in this event even, uses the term "machine learning" for some features (like some of their image processing stuff) and "AI" for other features. Their usage of the terms more or less matches my line of features I want and features I don't want.

Actually if Siri ever worked when I really need it, they could finally catch up with their promises made in 2011.

Semantically, AI has always been a superset of ML. So it's always been correct to call machine learning AI.

All machine learning is AI, not all AI is machine learning.

Yet there's a very clear distinction between when companies use the term "AI" and when they use "machine learning".

Not to be too, um, dismissive, but one of the things we discussed in my 300 level class called _Artificial Intelligence_ in college 2 decades ago was regular expressions, so, that ship has sailed far over the horizon.

Yeah, the annonying movies IPhone does from my photo library is something I'd love to opt out of. I get that some people love this feature but I don't. And as you say, I could ask the phone to do that for me on request.

Right? Siri lets me down on a regular basis - and has for years.

Apple has so hardwired my neurons to expect "It just won't work" when it comes to Siri that if I cannot physically reach my device, I just don't bother anymore.

Disable “show featured content” in Photos.

Well, yeah, Apple is being reasonable now because Apple just got through a big bad PR thing with their recent failed attempt at "AI". Apple are currently trying, as much as possible, to avoid applying the term "AI" to anything.

But that's not true of any other actor in the market. Everyone else — but especially venture-backed companies trying to get/retain investor interest — are still trying to find a justification for calling every single thing they're selling "AI".

(And it's also not even true of Apple themselves as recently as six months ago. They were approaching their marketing this way too, right up until their whole "AI" team crashed and burned.)

Apple-of-H2-2025 is literally the only company your heuristic will actually spit out any useful information for. For everything else, you'll just end up with 100% false positives.

Indeed, a good AI Snitch runs in background on your phone. The era of intelligent spyware.

I love this feature on both Photos and Google Photos

It's generally perfectly safe to ignore products whose main purpose is to ride some hype wave.

I don't think it's marketing's primary purpose to be off-putting?

Anyway, it's not the same thing: I'm fine with machine learning to give me better image search results, I'm not fine with machine learning to generate "art" or machine learning to generate text. Everyone has collectively agreed to call the latter "AI" rather than machine learning, so the term is a useful distinction.

it's a misleading distinction that's causing people to spiral out and think they're talking to an actual intelligence and is also being used to bamboozle lawmakers into allowing massive amounts of content theft

Who said anything about the product's purpose? Excessive use of the term "AI" is a decision of marketing departments, and happens entirely downstream of product design.

The same product could be produced five years ago or today, and the one produced five years ago would not be described as having "AI features", while the one produced today would.

(You can check for yourself: look at the online product listing for any mature "smart" device that got a new rev in the last three years. The Clapper would be described as an "AI" device today.)