While Apple keeps making money from overpriced hardware, the competitors are working on actually being pioneers in AI. It makes me sad to see so much computational power in my iPhone and iPad getting wasted on silly subpar iOS apps.
I have no doubt that Apple are working on local LLMs and combining them with Siri, it's only a matter of time. I expect we will see something next year, probably tied to the next iteration of hardware - they do want to sell hardware after all.
Apple have indicated that are working on this stuff, subtly. They avoided saying "AI", instead going for "ML", "encoder/decoder models" and "transformer models". The spell check on the next iOS is based on a transformer model. It's packed with image models for extraction, lighting and image correction. iOS is littered with ML stuff, that have "Nural cores" in their silicone.
This stuff is coming from them, when it's ready.
What I expect they will launch with a "LLM Siri" will blow our minds, it will be an all encompassing personal assistant. It will have linkage of all our documents, email, messages, movements, browser history, all locally securely on our devises. Not in the cloud. It will be so close to appearing like AGI that some people will clams it is.
Apple maps would have a laugh at that
The Apple Maps has better directions for my city at least. The Google one re-directs me through side streets and weird turns into traffic crossing lanes at the last minute.
Google Maps is loaded with suggested content. What about lunch? Need a haircut? Check out these top rated places!! Would you recommend Google Maps to a friend or colleague?
Apple Maps is like what Google Maps was years ago: a map and a search box.
https://www.cnet.com/tech/mobile/apples-worst-failures-of-al...
No one gets product planning right 100%. The key is to recognize failures early and cut them off.
MobileMe was quite a big misstep too.
not the first portable mp3 player
not the first smartphone
nor the first touchscreen phone
not the first bluetooth headset
not the first tablet computer
not the first smartwatch
not the first VR/AR headset
- The Lisa was so expensive that it failed to find a market regardless of how visionary it was, ultimately making less of an impact than the Apple II.
- The first iPod models were plagued with battery issues and had no meaningful way to replace them once it went bad (this extended into smaller models like the Nano).
- The iPhone was notoriously gimped at launch and only barely delivered on it's promises, with the majority of features people know today being added in updates.
- The first generation Airpods should be classified as instruments of sonic and physical abuse, not headphones.
- The first generation iPad was depreciated almost immediately and got ~2 years total software support.
To say nothing of their success, sure, but Apple clearly has their own honeymoon phases to work through.
Airpods Pro are much nicer, but should be the standard. The Airpods 1 (and frankly the Airpods 2 as well) sound thin and are borderline unacceptable for what they cost. Just my $0.02 though, audio and comfort are ultimately subjective.
Yeah, Apple would want to have voice chat, not text chat. Especially on a phone
This was recently submitted to HN and I think it's largely reasonable supposition:
All text within photos is being automatically recognized and is fully searchable. That has saved my ass on occasion when some piece of information was in a photo.
The AI background removal feature in Photos is cool too — just drag from a foreground element, and it usually gets it right.
But despite this kind of flawless feature integration work, clearly Apple as a company has some issues with applying generative models on a more fundamental scale. They look a bit stuck in their own loop of trying to replicate past success with an upcoming $3,500 device that’s all about ever fancier glass and aluminum hardware.
Qualcomm is announcing hardware that won’t be available until (at least) next week. That’s just not Apple’s way, and I think it’s a mistake to say they’re ignoring what’s going on.
They weren't silent about AI child porn scanning on-device so we know they definitely have the capability for AI on device.
Contrast that with the rest of the industry who lead out with "We're doing AI!" and then they connect it to this and that feature.
I'd argue that consumers care about having great features, not about what particular technology powers it. My phone could run on unicorn poop, and I don't care as long as it does what I want it to do.
The power of LLMs is that they can allow a human to interact with a computer using natural language.
Imagine just asking Siri to order you something and Siri finds the cheapest legit supplier and just orders it from them. It would remove the first hop needed to go to Amazon and unlike Alexa Siri is always with you. Google and Apple could do to Amazon what they did to Facebook.
And they are making tremendous investments in Metal development.
Also, there are native apps utilizing both Stable Diffusion and Llama... And they are buried under a tremendous mountain of other AppStore junk.
The large gap between Apple's demos and backend development and actual deployed ML apps is very unfortunate and sad, but its not uncommon either.