They have already probably optimized it as much as they possibly can while still maintaining the accuracy of the output. Apple has released lots of research papers towards that, and they have released some AI models on huggingface if you are curious about that. Quantization is one method used to reduce the size of the AI models.
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 Intelligence' Personal AI Unveiled for iPhone, iPad, and Mac
- Thread starter MacRumors
- Start date
- Sort by reaction score
Since when are iPhones, iPads and Macs free?And it’s free. When a product is free, you’re the product. Or does it not apply here because
Looks like our forum has become very 2024 - fully crammed with negativity.
I'm currently installing the beta on my iPhone so I'll tell you in a bit. The automatic sorting thing and so in Mail I think they just called ML and it's not behind the Apple Intelligence moniker. Scheduling messages etc. is coming to all. Probably what they mentioned after introducing it as Apple Intelligence is the 15 Pro exclusives. The rest of us get an uglier control center and overcomplicated Photos UI.
Not really. Seems mostly powered by Apple with a little ChatGPT duct taped on for trivia questions. And you get to choose if it uses ChatGPT or not.
Not trying to troll, but that often is how chip designs work.
Fabs send their clients a PDK. It has information about Mtr/mm2, yield, PPA, as well as other information on yield/performance metrics given a reference design.
Custom Chip designs are done using the information in the PDK, but deciding to increase your core size or whatever isn’t technically hard. The hard part is convincing your manager to eat the cost of a higher core size, not telling the fab what dimensions the core will be.
Where does it say that Apple Intelligence is repackaged ChatGPT? Because from what I saw it is Apple software running on Apple Servers.
They probably have several. There are many LLMs available with different Parameter sizes. For many purposes you do not need a full fledged model with billions of parameters that require loads of ram and compute power. Like a Flan-t5 with fine tuning. ChatGPT-4o cannot run on a 8GB device anyway (no way). So they will have purpose built models with less parameters for certain tasks and a more complex one for Siri for general understanding. Which will still be less powerful than ChatGPT as it has to run on device with limited compute power (and most of all - memory). For more complex queries you will need to offload.
I'm not talking about windows security; I'm talking about how Microsoft secured and stored the index data specific to the recall feature. When researchers get hands on iOS 18, they can also figure out how secure the index data are for the AI feature on iOS 18.
It’s the entire fabrication process. Just having an EUV machine doesn’t guarantee anything. One of the ways TSMC and Samsung solved low yield on 4nm is through a technique called yield-capture that has nothing to do with EUV. There are thousands of steps in a chip fab and the EUV is a critical but small role in the overall process.
Apple’s role is insignificant.
Also, yes, those OLEDs are Samsung/LG-developed displays with Apple having no input in its development. It’s literally a Samsung M-panel display set technology in most of the pro models. You ever wonder why Apple can’t just go to one of those hundreds of Chinese display makers to get those panels? It’s because Apple doesn’t develop the tech in their displays. Remember tandem OLED that Apple bragged about? Turns out Dell also has it now in their laptops because they source it from LG, who developed it.
Last edited:
Not really. The first thing I saw them do with Siri was ask what’s the weather like somewhere else. It went downhill from there.
Then they mentioned ChatGPT and things got better. Except for this annoying little prompt is it ok to use something smarter than Siri (which is basically any third party ai service).
I found it very exciting and might actually install a beta this year. You can sit this one out.
Omg im so worried about this and privacy :/ i hope you can turn it off
What makes you think it’s not their own AI technology? Most of what they presented today is their own tech. The ChatGPT component is just bolted on to handle some specific task that take special handling or require web results like trivia questions. You might want to rewatch the presentation.
The whole compose and rewrite feature is ChatGPT.
Apples intelligence? Just more of the same we’re used to. Anything significant will be handed off to ChatGPT.
I understood the same. Apple Inteligence has nothing to do with Chat GPT. It seems the only thing they did with chat gpt is to put it as an option when you use Siri.
Maybe they will use the OpenAI models for the rewrite feature. But the presentation was pretty clear, that compute is done on Apple servers.
Did you read this with your own eyes, or did you wear glasses?
Really, it’s just another tool.
Ok stand still, we will contact emergency services and have the gun pointed at your head forcing you to use this removed