Apple Intelligence news summaries are back, with a big red disclaimer
- Thread starter JournalBot
- Start date
I will never understand summarizing headlines, or notifications in general really. It's a headline, its already summarized. What you are doing, reducing it to a single word?
Upvote
197
(200
/
-3)
Exactly! If AI is going to do something then how about use AI to analyse when something important is happening based on content and volume, and then choose the most relevant/trusted/headline/source to display? You don't have to change it, there's no value in that.
i.e. if I'm in TownA and news sites in/referring to TownA are all broadcasting a tornado warning, then maybe display me the relevant headline from the meteoric service that covers my area?
Upvote
58
(58
/
0)
Going with "if you can't fix it; mark it 'beta' and call it good" doesn't seem like Apple's historical MO; but 'AI' seems to have a reality distortion field that you'd need an astrophysicist to reckon with; so I guess it bends everyone eventually.
Upvote
89
(90
/
-1)
I'd be very interested to know how the incentives shake out; but there seems to be fairly limited interest from mobile OS vendors in trying to deal with the fact that notifications quickly become a hellscape of slop unless you turn almost all of them off.
At least attempting to offer spam filtering/prioritization on-device to deal with apps that are...liberal...in their interpretation of what notifications are actually important would be a genuinely interesting feature. I can only assume that saying the quiet part out loud about how most notification are garbage would upset some sort of services revenue equilibrium, though, especially when even Apple can't resist the siren song of burning goodwill for quick monetization.
Attachments
Upvote
36
(38
/
-2)
An app can send multiple notifications, the summary feature will attempt to summarize all/most of them, grouped by thread or app (depends on app and your notification settings)
Last edited:
Upvote
-3
(6
/
-9)
We already have inaccurate summaries of news articles; they're called headlines.
Upvote
87
(88
/
-1)
Some people laugh at me when I’m telling them I’m still using RSS. Still the best imho: have headline, work offline, privacy ok, can save article and more.
Upvote
54
(55
/
-1)
I remember getting this notification and literally writing a complaint to Apple about it. Notifications were supposed to be for things I need to know, not another free advertising space for every money grubbing company.
Apple can now consider themselves part of the elite club, alongside Microsoft, of companies who are filling their OS with spammy ads.
Upvote
54
(55
/
-1)
or just leave that shit off and read the article(s) yourself.
I still have Apple Intelligence disabled and will continue to have it disabled. It's just a waste of battery power for a, possible, incremental benefit.
The couple times I've turned it on I just can't find a use for it at all. I don't need summary of incoming texts or summarized reminders of my day or anything summarized and the standard siri functionality for kitchen timers or turning on a music station or a movie on the TV is all I ever use Siri for. Maybe do a conversion for baking / cooking while my hands are dirty. It doesn't need AI to do that.
The one cool feature Photos has is going through 10 years of photos and finding all picture of person X or dog B or cat F and letting you put them in a "labeled" album. Unfortunately, if you remove that tag there doesn't seem to be a way for Apple to force "rescan" it. That is...it was putting pictures of my parents old Dog X into the alblum for our new Dog Y because they look very similar (same breed) and when I was moving Dog X into his own album I removed a few tags vs. a rename and now I can't get that handful of pics to actually be recognized as that album.
I still have Apple Intelligence disabled and will continue to have it disabled. It's just a waste of battery power for a, possible, incremental benefit.
The couple times I've turned it on I just can't find a use for it at all. I don't need summary of incoming texts or summarized reminders of my day or anything summarized and the standard siri functionality for kitchen timers or turning on a music station or a movie on the TV is all I ever use Siri for. Maybe do a conversion for baking / cooking while my hands are dirty. It doesn't need AI to do that.
The one cool feature Photos has is going through 10 years of photos and finding all picture of person X or dog B or cat F and letting you put them in a "labeled" album. Unfortunately, if you remove that tag there doesn't seem to be a way for Apple to force "rescan" it. That is...it was putting pictures of my parents old Dog X into the alblum for our new Dog Y because they look very similar (same breed) and when I was moving Dog X into his own album I removed a few tags vs. a rename and now I can't get that handful of pics to actually be recognized as that album.
Last edited:
Upvote
37
(38
/
-1)
Sadly a lot of headlines are clickbait and uninformative. I try to stay away from news sources that go this route, but I'm not iOS's typical user.
Upvote
-17
(6
/
-23)
This strikes me as treating the symptom rather than the disease, like a lot of the use cases people try to offer for these AI tools.
If you're getting bombed with too many unhelpful notifications, you need to limit what can send you notifications.
If you're getting bombed with too many unhelpful notifications, you need to limit what can send you notifications.
Upvote
40
(40
/
0)
"This summary may be wrong, verify by reading the original text"
Very useful tool
Very useful tool
Upvote
64
(64
/
0)
I’m still wondering who asked for any of this? In general, an article lede provides the info needed for me to know whether I want to read further - a poor summary of the lede isn’t gaining me anything.
Upvote
26
(27
/
-1)
Billions and billions of dollars and AI can almost accurately summarize headlines. This is a huge win. I often have trouble reading a headline and getting a brief summary of what may be in a headline will really help. Its hard to imagine how we ever made do before!
Upvote
48
(50
/
-2)
It's for people who have developed super short attention spans thanks to platforms like Tiktok. I swear to god anything longer than a single, easy to digest paragraph is too much for a lot of people nowadays.
Upvote
11
(14
/
-3)
At least Apple are providing the option to disable this, which I do. Haven’t missed it as another commenter noted.
Upvote
15
(15
/
0)
I wouldn't have expected this from Apple. Why not be the adult in the room and say NO to an idea that hasn't been fleshed out? A blemish on their reputation.
Upvote
22
(26
/
-4)
tl;dr; Apple: "our stuff is still broken, but you've been warned" (walks into the sunset with triumphant background music)
Upvote
15
(20
/
-5)
The answer is simple: you overestimate current AI.
It just can't do it. At least not reliably, and within budget.
But even if you're skeptic, the hype machine has still dragged you into the mindset that "surely it must be easily able to do THAT, at least".
Upvote
22
(26
/
-4)
chaos215bar2
Ars Praefectus
Or implement actual spam filtering, limited time notification enablement requests, etc.
Notification spam has gotten so bad that apps will literally block in-app functionality if you don't enable notifications. I've seen an app hide delivery status and updates if you don't enable notifications (even directly in the app). Google these days will force you to install YouTube and enable notifications for "account verification" — but only to do things that don't immediately make them money, like removing an unused payment card.
Notification spam is really bad and should have frankly been a blindingly obvious problem to see coming. Unfortunately, rather than using their position as the gatekeeper of the App Store to do anything about it, Apple is busy seeing if they can get in on the action.
Upvote
18
(18
/
0)
A user can uninstall an app or turn off notifications and any app that sends multiple notifications that need summarization sounds like a prime candidate for removal.
Upvote
9
(10
/
-1)
I disable notifications from like 90% of the apps I have installed. I think a handful of banking apps (mostly to see possible fraudulent activity ASAP), iCal and iMessage, a couple of other Apple native apps are it. I disable notifications or just our right delete a lot of the Apple native apps I don't use as well. All the rest...I don't need to be "notified" when something changes. If I want to know where an order is or whatever I'll just go look in the app. I don't need to be pestered with info that can wait a day or doesn't matter.
Upvote
31
(31
/
0)
One spam notification, and the app gets notifications turned off.
No exceptions. Not even “messenger” apps.
Upvote
25
(26
/
-1)
This is an especially bad use of AI summaries. The first thing you read or hear about a subject is more likely to imprint than any subsequent retraction, correction, or new information. Specific situations will vary, but across a broad enough group of people, this holds true far more often than it does not.
In this age where a disconcertingly large portion of the western world is living in an alternate news reality, why would anyone think it's a good idea to risk the creation of even more confusion?
In this age where a disconcertingly large portion of the western world is living in an alternate news reality, why would anyone think it's a good idea to risk the creation of even more confusion?
Upvote
17
(18
/
-1)
trash. who needs a summary that i need to verify every time? as if anyone would do that.
Upvote
12
(13
/
-1)
Upvote
3
(3
/
0)
“We summarised an entire book into a five page précis. We then reduced that to a handful of paragraphs. These were boiled down to a few pithy sentences. From that we produced a single short sentence to give the essential nature of the work. Finally we realised that we could concentrate even that down to a single word.
We had reduced an entire book down to a single word!
But then we forgot the word.”
Upvote
16
(16
/
0)
This. When faced with reading even a short article, many people will just avoid it completely or complain it’s too long. Yet these same people will happily watch a thirty minute YouTube video that line by line walks them through a written article that would take three minutes to read, complete with additional advertisements and bias added. It’s pure madness, yet they don’t see it that way.
Upvote
12
(12
/
0)
Why are people downvoting this?
Clearly the incentive for the publisher of the notification is not to give everything away in the title, but to provide just enough info to make people click. Having notifications that actually summarizes the content would be a very useful feature. Preferably it would be done by the human writing the article, but that’s not the world we live in.
Upvote
-6
(3
/
-9)
When you start putting disclaimers like this out it suggests to me that none of the fundamentals have changed and it still doesn't work. No one wants to admit that the AI bubble will probably be the most disastrous era of technology in 50 years.
Upvote
9
(10
/
-1)
Afaik, this is feature summarizes the content, not only the headline. If it worked as advertised it would be a useful feature in the world of purposefully vague and misleading click bait headlines. This is of course not meant to replace reading the article, but to provide a quick overview of what’s happened over all your sources suitable for the lock screen.
Upvote
-3
(3
/
-6)
Agree. Apple's proposals for "Apple Intelligence" were interesting... But unsurprisingly, not deliverable. My bet is they will never be able to deliver it to their standards with LLM tech, and they should have known better. But the distortion field is too strong.
Upvote
10
(10
/
0)
Really don't get this. What is the use of a summarisation feature that does not (always) work. It apparently isn't correct enough for the user to be able to rely on its accuracy, so that means it's just getting in my way / creating extra confusion. What is the point?
The same goes for the AI answers on search engines like Google. Often, I find the "answer" is completely opposite to even the first few reliable hits on the first page of search results.
Gen AI has uses when generating new content, and has clear and serious flaws when it comes to interpreting and summarising information presented to it.
Yet, here we are, being seduced to rely more and more on shiny tools that cannot even uphold their most fundamental contract with the user: To provide an accurate representation of information of interest.
Meanwhile, we are becoming ever more reliant and lazy, losing essential skills of interpreting, analysing and weighing information. The "dumbification" is accelerating.
The same goes for the AI answers on search engines like Google. Often, I find the "answer" is completely opposite to even the first few reliable hits on the first page of search results.
Gen AI has uses when generating new content, and has clear and serious flaws when it comes to interpreting and summarising information presented to it.
Yet, here we are, being seduced to rely more and more on shiny tools that cannot even uphold their most fundamental contract with the user: To provide an accurate representation of information of interest.
Meanwhile, we are becoming ever more reliant and lazy, losing essential skills of interpreting, analysing and weighing information. The "dumbification" is accelerating.
Upvote
6
(7
/
-1)
I find it a very intriguing trend that right now, generative AI and specifically LLM's, are presented as a source of factualness in all sorts of tools and helpers, where its very purpose is essentially to "make up"/generate stuff through statistical inference. The AI's job is not to remember and reproduce "facts" but rather to come up with an answer that is so resembling of the truth (or coincidentally actually happens to be the truth) that it is hard to tell a truth from a hallucination. This flaw (or feature, depending on how you look at it) is very fundamental to the way LLM's work, and not simply a bug to be fixed in the future. By their very nature, LLM's can only provide answers that might be true (even though the degree of confidence can obviously always be improved given enough data and compute).
It is a bit like asking a generative image model to "create" an actual photo of something real. It may come very close, but it may also be completely off because it is not its job to be an image archiver, but rather to be an image generator.
It is a bit like asking a generative image model to "create" an actual photo of something real. It may come very close, but it may also be completely off because it is not its job to be an image archiver, but rather to be an image generator.
Upvote
12
(14
/
-2)