So basically Apple found a way to create a new feature which just happens to thrash your iPhone's storage device into premature failure and thus drive additional sales. Cute.
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 Develops Breakthrough Method for Running LLMs on iPhones
- Thread starter MacRumors
- Start date
- Sort by reaction score
I really do hope Apple is able to make remarkable progress in the realm of AI. Its one of the few companies that I trust my data with.
I suspect, by the time write induced failures would be expected to occur the iPhone would have long since recycled.
Interesting. As Siri is still pretty much unusable, every upgrade to her brain is welcome.
And it’s not just the voice assistant but also every other Siri smart thing they have added over the years: from shortcuts, calendar suggestions, smart data detectors, … it’s all so unreliable and insignificant, I bet most users don’t even know about these features.
And it’s not just the voice assistant but also every other Siri smart thing they have added over the years: from shortcuts, calendar suggestions, smart data detectors, … it’s all so unreliable and insignificant, I bet most users don’t even know about these features.
Last edited:
I don't want LLMs, I just want coffee.
Is this really a breakthrough? The latest iPhones have as much RAM as the starting MacBook Pro. I wouldn't call it "limited". And a RAM disk isn't a breakthrough either...
Also, we're constantly hearing about how only the newest Apple products can support their latest technology, so I suspect we will still see this model limited to push people to upgrade.
Also, we're constantly hearing about how only the newest Apple products can support their latest technology, so I suspect we will still see this model limited to push people to upgrade.
Last edited:
Most phones on the planet don't have that much ram. Not to mention this could be used on MacBooks with 8GB of ram. So everyone gets to use it.
This is something I tried to point to when people say on Macs that 8GB is enough. Well not specifically this, but some new feature that will be more RAM intensive that will struggle earlier for those people who have the base 8GB configuration that Apple has stuck with for so long.
Odds are we can live with significantly less for a version 1 that lets us actually integrate the AI with the OS in a way that is useful. I would prefer to tell my phone to change specific app settings or something than go through settings to find it.
Perhaps it's my age, or perhaps it's my life experience, but I don't want LLM-AI in my life. It's too dependent on the perspective of its trainers. But I will let it start my coffeemaker in the morning.
What most people don't understand about Siri (or even Apple design) is that Siri doesn't give you what you want. She gives you what you need. She's not so much an assistant as she is a wise omnipotent technological resource. Next time she doesn't answer the way you expect her to, think about why, go deeper than the surface of what she said. You will likely find something deeper than you could have ever imagined.
This is very neat research, but I think the suggestion from the headline, that this will bring LLMs to iPhones, is a stretch. The test platform was a M1 Max with an unspecified amount of DRAM. My educated guess is that it had at least 16GB of DRAM, likely even more. This is not iPhone territory.
I said it before and will repeat it now. Making RAM a luxury with extremely overpriced upgrades will bite Apple, or rather us customers, in the ass.
I said it before and will repeat it now. Making RAM a luxury with extremely overpriced upgrades will bite Apple, or rather us customers, in the ass.
Last edited:
Apple doomed again. And again and again.
They must be squeaking by keeping the doors open, making all of these terrible decisions over the years.
It's the WRITE cycles that quickly degrade SSD drives. Reading is orders of magnitude less problematic. Almost all of the bytes cached in this application are static and only written once to flash. I don't see a problem here.
Modern SSD chips have incredibly long lifespans. The SSD in your Mac takes more of a thrashing in its normal daily operations than this would cause. They're designed for this.
Dont come for my girl sir. LOL.
She tries....her best
Siri reminds me a lot of my best friend i have had since we were 2 (we are 27 now). Reliably useless for most important task but for menial task (for siri playing a song and for him, always with the whiskey on hand) they are very useful lmao
Agreed. My wife is Chinese. I speak the barest amount of Chinese and she speaks passable English, but she has a really hard time reading English. When I know she's too busy to manually translate my texts, I manually translate my text to Chinese. I cannot for the life of me understand why Messages doesn't let me assign an "auto-translation" to certain people, as in, when I type a text in English to that particular person (my wife) ALWAYS send it to her in Chinese. Seems super duper easy peasy. But, nope. She could do the reverse, of course, where whatever she types or dictates in Chinese automatically translates to English if she sends it to me. Seems obvious as hell to do this.
HobeSoundDarryl
macrumors G5
So you’re saying I should bake that cake just 15 minutes instead of 50? Not so sure about that, but I’ll give it a try.
"Apple researchers have developed a novel technique that uses flash memory – the same memory where your apps and photos live – to store the AI model's data."
This sounds similar to what Samsung has on the S23 Ultra, where up to 8 GB of storage can be used temporarily for memory.
This sounds similar to what Samsung has on the S23 Ultra, where up to 8 GB of storage can be used temporarily for memory.
HobeSoundDarryl
macrumors G5
Apple: "Our parameters are like 10- er, I mean- 15 times competitor parameters."
Fans: "Apple may not have the most parameters, but they utilize only the right ones"
Both
Last edited:
Hey Siri, did you write this post?