Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Apple's First AI Features in iOS 18 Reportedly Won't Use Cloud Servers
- Thread starter MacRumors
- Start date
- Sort by reaction score
The Encyclopedia Britannica used to be available on CD-ROM and later DVD. As a corpus of information and curated facts to draw on without the need for an Internet connection, this was likely an invaluable resource for many people of all walks of life interested in gaining knowledge. The problem was that as an offline resource, it could never stay current or include any new information beyond its publication date.
If Apple can manage to condense that kind of knowledge resource into an on-device database that can then be accessed offline using an AI-driven speech engine that can process English or any other source language in a fluent and natural way, I imagine that might go a long way towards curbing the tendency people have nowadays to simply throw it to ChatGPT, Gemini or Bing Copilot and take their answers verbatim.
Of course, encyclopedias and generative AI are two different things, but they're both built off a corpus of knowledge. If Apple can license a well-established knowledge base like Britannica and pair it with the right technology that lets people actually do something useful with it offline, at least this would help avoid the copyright problems plaguing generative AI now.
"Putting Wikipedia on an iPhone" (and in multiple languages!) seems prohibitive from a storage point of view, but if that's one of the directions that Apple is going for, it at least seems consistent with their environmental and educational goals.
Of course, maybe I'm reading too much into Gurman's take on this, and Apple's new AI engines are still going to search for knowledge online, just make it easier and more natural for you to do it (i.e. a revamped Siri).
If Apple can manage to condense that kind of knowledge resource into an on-device database that can then be accessed offline using an AI-driven speech engine that can process English or any other source language in a fluent and natural way, I imagine that might go a long way towards curbing the tendency people have nowadays to simply throw it to ChatGPT, Gemini or Bing Copilot and take their answers verbatim.
Of course, encyclopedias and generative AI are two different things, but they're both built off a corpus of knowledge. If Apple can license a well-established knowledge base like Britannica and pair it with the right technology that lets people actually do something useful with it offline, at least this would help avoid the copyright problems plaguing generative AI now.
"Putting Wikipedia on an iPhone" (and in multiple languages!) seems prohibitive from a storage point of view, but if that's one of the directions that Apple is going for, it at least seems consistent with their environmental and educational goals.
Of course, maybe I'm reading too much into Gurman's take on this, and Apple's new AI engines are still going to search for knowledge online, just make it easier and more natural for you to do it (i.e. a revamped Siri).
so I have a little anecdote to share about the moment I realized AI could actually be useful for photo editing. was looking through photos & spotted a lovely picture of someone very dear to me…with someone I don’t quite care for in the frame, just kinda blurry. was thinking “wow, would be nice to just have that person not be in the photo…” and had a little moment of envy for Pixel owners.
I’m not gonna act like I’m not over the hype of generative AI, it’s entirely too much. ChatGPT/DALL-E are not the solution to the world’s issues. that being said- I do think there is a place for such technology. in an age where Photoshop has existed for decades, why not automate simpler tasks for the average user? we as humans are constantly looking for ways to automate, for better or for worse—I think image editing through generative AI is a change for the better. image creation gets into an entire debate that I’m not gonna bring up in a non-PRSI thread.
also, if it wasn’t obvious, I’m an extremely verbose person and cGPT does occasionally help me to make important emails/texts more concise. I do find that helpful. 🤪
AI is becoming a devalued word like form factor, price point, and use case.
Are you sure about that?
Turn off cellular.
Turn on airplane mode.
Turn off WiFi.
Ask Siri who was president in the year 2000.
Siri: “To do that you need to turn off airplane mode.”
Turn off airplane mode.
Ask again.
Siri: “To do that you need to be online.
NFT 2.0 is it. All across the internet many incel and far right profiles embrace AI everything, plastic cartoons replacing apes. A meme factory for the disinfo addicts/pushers and H*tler apologists. The crypto 3.0 youth burned by their losses or emboldened by their wins exist in a community weaponized to finance and help further push unreality. So the muddying of the waters caused by AI infecting shared truth is no accident.
When shared truth/news is made unreliable/unkowable, we’re left with nihilism, fractured bits of truth diluted with lies, or the propaganda of the powerful writing reality. Humans can only handle so much anxiety, so many will retreat from caring. I’m not saying this is a new thing, but the advent of LLMs and AI for text and images and video endangers humanity at a new scale. Watching gamergaters become cryptos become NFTists become AI-enthusiasts has not been pleasant.
Strongly disagree. Photos are over processed under some circumstances, but not all or even (in my experience) a significant minority.
Kinda wild to see people think generative AI is just what they see with ChatGPT.
This is exactly what I want to know. All of this talk of the M4 series of chips makes me concerned about the capabilities of my lowly M2.
Completing sentences is so lame if that’s a headlining feature.
Just give me a Siri that works with a modicum of intelligence. It’s so embarrassing for Apple that Siri is not much better than what Apple introduced in the 90’s with the Quadra AV models.
Just give me a Siri that works with a modicum of intelligence. It’s so embarrassing for Apple that Siri is not much better than what Apple introduced in the 90’s with the Quadra AV models.
I’m just mulling over how much this would cost as a subscription. I can’t see Apple doing it for free.
Totally agree!
The only use I'd make of AI is asking trivia questions, and (hopefully) get a real answer, or give vocal commands to my devices.
I think a lot of people share your fear, including me. I bought an iPhone 15, and eventually not getting a chatgpt-like feature "just because of lack of power" kinda makes me nervous and anxious.
Probably any iCloud subscription will do. At least i hope
I'm fascinated to see what can be done with local AI on a device with low RAM. Either the output with be low quality or incredibly slow. I can't see how they'll get around that hurdle.
There's a LOT that's not cloud-based. Anything llama.cpp, ollama, mistral, and the list goes on... Some requires large amounts of RAM, some models are optimized for smaller amounts of RAM but suffer as a result. Those can probably be fine-tuned for specific tasks however, which is probably what Apple is going for (and what they've already done for text completions in iOS17 for example)
I know what you meant, but there's really no bottom that discussion. It's turtles all the way down. Best you can do is have a photo that reasonably approximates your perceived reality.
That’s what a lot of the user facing hyped services are doing, yes.
What it’s really good at is being an expert system. If you can train it on just the data you care about, it’s quite good. It’s what we’ve been using computers for in the first place. We have the data, we know what we want from it, we want the computer to process it for us.
I do share your lack of optimism that it’s going to become a truly useful consumer service any time soon due to the issues highlighted. It’s like the web and all the other technologies. Great potential, but commercialization ruins everything even as it enables it to exist in the first place.
This already happens in iOS but has historically been called Machine Learning (ML). It’s this feature that allows you to search for cat pictures in your Photos library, for example.