Apple's AI moves are nothing new for a company that has integrated outside tech for decades.
See full article...
See full article...
Apple and OpenAI currently have the most misunderstood partnership in tech
- Thread starter JournalBot
- Start date
Seems like a huge gamble to commercially integrate and deploy these so quickly and on such a large scale. Large Language Models are advancing by leaps and bounds all the time, but I really don't think they're ready for prime time just yet...
Upvote
24
(41
/
-17)
- Add bookmark
- Featured
- #6
Benj, thanks for bringing some context, nuance and sanity to this. If the Ars comment threads are any indicator, a lot of people have poor/kneejerk understanding of both Apple and AI. Maybe this will help make discussions more thoughtful and interesting.
Upvote
99
(120
/
-21)
- Add bookmark
- Featured
- #7
Elon's bizarre freakout on the subject has been amusing to watch, especially because it's very transparently due to his animosity towards OpenAI. Sorry Elon, despite whatever misgivings people may have about Apple integrating ChatGPT, literally nobody would have preferred integration with Grok.
Upvote
153
(157
/
-4)
Pushing the Apple Intelligence branding upfront gives them the space to swap to their own models when they feel they're ready.
Upvote
108
(112
/
-4)
Many people (annoyingly) talk on their phones to other people like that. They're usually the people who think 'speaker' is the only mode for phone calls or audio experiences with their phone.
Upvote
84
(85
/
-1)
Using all of his mighty authority, he has decreed that if you visit a SpaceX or Tesla facility and you have an iPhone, you have to leave it up front. Way to go, Elon! Really stickin' it to 'em. That'll show Apple what's what!
Upvote
39
(43
/
-4)
- Add bookmark
- Featured
- #11
This is single-handedly the best article about the announcements’ impact I’ve read. Far far better than the shit-takes at the WaPo and NYT, which seemed to amount to ads for OpenAI.
Upvote
56
(62
/
-6)
OrangeCream
Ars Legatus Legionis
You can ask people to pay more for a feature if they don't know the feature exists, after all.
It's been wildly successful, too; they've had E2E encrypted messages since 2011. How do you know that your chat is E2E encrypted? Blue text bubbles.
Of course it helps that the competition, Google, didn't offer E2E encrypted messages until 2023.
Face Id is another one. They offered it in 2017 with the iPhone X. Even in 2022 it's not a headline feature of Android, because it isn't secure (enough) by default:
https://www.xda-developers.com/goog...back-face-unlock-but-more-secure-than-before/Unlike the Pixel 4 series, which used a a miniature radar camera system (Soli) to grab a 3D map of our faces that's similar to the iPhone's approach, the Pixel 7 and Pixel 7 Pro's face unlock will just use the single 10.8MP front facing camera.
Of course individual manufacturers have implemented 3D face unlock:
https://www.phonearena.com/reviews/Huawei-Mate-50-Pro-Review_id5624
- 3D ToF sensor for face recognition
Upvote
31
(48
/
-17)
I'm amusing myself with the thought of some brilliant key engineer on the ragged edge of burnout from putting in 15-hour days at SpaceX who is currently writing his resignation letter because the increasingly unhinged maniac ostensibly at the helm of his employer has thrown a knee-jerk temper tantrum threatening to forbid iPhones at his company.
Upvote
49
(53
/
-4)
Is there anything outside of Elon's toybox that uses Grok? Seems like it's just another 'pay attention to me' thing.
Upvote
53
(56
/
-3)
Upvote
124
(132
/
-8)
Having grown up when phones were tethered to the walls, it's always startling for me to see someone talking like that to their phones.
I always get the impression they're eating pizza...
Upvote
51
(51
/
0)
At least 3DTV never handed out toxic pizza recipes or trip itineraries featuring places that don't exist. It's got that going for it. Seriously though, the potential liability here for Apple seems enormous. But I guess if any company has the war chest to wade through a vast legal onslaught over something like this, it's them.
Upvote
13
(21
/
-8)
I've started loudly singing right at them when I see people do this, making up songs with lyrics like "IIIIIIII am a fuckwit who uses speakerphone in puuuuuubbbbbliiiiiiiccccccccc"
People other than the idiot on the phone usually enjoy it!
Upvote
36
(43
/
-7)
The obvious question is: is your data safe(r) with any Musk company or with Apple?
Upvote
19
(19
/
0)
If my CEO tried to force everyone to switch to android and windows, I'd absolutely quit just for that alone. Maybe not on the spot because I could easily just do minimal-productivity-to-stay-employed while I find something else.
It's not that I'm a hardcore platform elitist, I'm just not a fucking idiot and don't want my work to get more difficult because of a window licker's temper tantrum
Upvote
62
(65
/
-3)
Partnering or (especially) acquiring tech from outside is one of the most standard plays in tech. Do so many people really think buying a startup means history changed, and the startup’s tech was magically created in house?
”Ohhhh we were always at war with Oceana!”
”Ohhhh we were always at war with Oceana!”
Upvote
27
(27
/
0)
And now he can use the AI on his iPhone to rewrite his letter to make it more professional and tone down the violent rhetoric surely saving him from revenge prosecution. Win-win.
Upvote
32
(32
/
0)
Thank you for the kind words rmohns and vought122, I deeply appreciate it. I'm a big fan of nuance and resisting extreme black-and-white (all or nothing) thinking on topics whenever possible.
Upvote
92
(94
/
-2)
More or less. Well, that and a "literally pay me" thing--he's been trying to threaten that he'll refocus his AI development at Twitter/Grok and xAI instead of Tesla unless his absurd pay package is reinstated. Not that I find that a remotely compelling argument, but then maybe that's just because I'm not the target market. After all, I'm also not stupid enough to be a Tesla investor at its current valuation or given its current leadership.
Upvote
18
(18
/
0)
The stakes are higher. Not only is Siri far more limited and way less likely to go completely off the rails, but most people would probably be embarrassed to admit they did something stupid because Siri told them to, whereas people are already showing up in court claiming that ChatGPT led them astray as a defense.
This has been a common concern since at least the broad rollout of public GPS. At some point you have to know better than to drive into the lake, even if the computer lady seems to be telling you to do that. Nowadays, a lot of chatbot tasks are far more subtle and complex than "drive your car in this direction," so it can be much less obvious when the computer lady is wrong...
Upvote
12
(13
/
-1)
I've seen a lot of claims that "requests are not stored", and I believe them, truly. But I'd like to know if the responses of the model are stored? This would seem like important thing to know.
Upvote
17
(17
/
0)
Could you specify what that aspect might be? Since Apple has specified that part of their deal with ChatGPT is that queries are not stored on OpenAI's servers, concerns about shoveling data to OpenAI are unfounded (particularly since all such queries are opt in). The rest (that is to say 99%) of Apple's announcements are specifically designed around the idea of privacy and data safety. I think Apple has done a really interesting job of re-contextualizing "AI" as something that can leverage what your device already knows about you to provide actually useful functionality, but everybody just saw ChatGPT was mentioned and lost their damn minds.
Upvote
52
(56
/
-4)
Snarky Robot
Ars Legatus Legionis
You know, it’s funny. I always heard that if you posted something stupid and the community hid it, all you had to do was copy-paste and it would seem smart.
I think we have proof that isn’t the case now.
I think we have proof that isn’t the case now.
Upvote
18
(18
/
0)
But the way ChatGPT is implemented here is precisely the same way Bing was implemented in the past. Most of the way Apple is using LLMs are obvious improvements to Siri.
I feel like this article, while clarifying the situation, still leaves people with the impression that ChatGPT is going to respond to most Siri responses. If you ask Siri, "Hey, can you send this page as a pdf to my wife" and then "tell her I think she'll enjoy reading it," it can now figure out how to do all that and also who "she" is and switch pronouns ("she'll" to "you"). That's legit useful stuff that LLMs are great at
If you say, "how do I keep my cheese from sliding off my pizza" it'll say, "ChatGPT might help you with this" and you have to say, "yes, I want to go to ChatGPT for this."
Upvote
49
(49
/
0)
It's also just really obvious how fuckin' dumb Elon is. Literally, all you have to do to understand the things he's performatively whargarbling about is watch the fucking keynote. ChatGPT is an opt-in, optional extension for questions that exceed the Apple LLMs' scope and training data. It's not "integrated at the OS level," whatever the fuck he thinks that means, and the only thing that was, was indeed developed in house by Apple.
He's just......not an intelligent man. He has moments of cleverness, at best, and he has some facility with technical information, but he's not well informed, he’s impulsive and emotional, and he thinks he knows everything about everything, and that’s a killer recipe for a dumb tryhard who humiliates himself on the daily and isn’t even aware he’s doing it.
Upvote
60
(64
/
-4)
No, you don't have to trust OpenAI at all--the function is going to be explicitly opt-in with confirmation for every request, and (as the article mentioned) without deeper OS-level integration that could potentially leak data to OpenAI.
Upvote
31
(32
/
-1)
This was a helpful article.. I'm still a little surprised that Apple went with OpenAI, who is closely backed by Microsoft. Plus the OpenAI brand is "challenged" regarding privacy, which is kind of Apple's lane. I would have thought that from a brand-perspective, Anthropic's Claude suite of models would be much more suitable, and Anthropic's reputation on privacy seems slightly less toxic than OpenAI. So that's my main confusion - why not pick Anthropic over OpenAI as a partner? Their models at this point are quite comparable on quality, though gpt-4o is really good on latency/response time.
Upvote
11
(12
/
-1)