I hadn’t really even considered that apple wouldn’t be working on their own LLM. Seems like everyone is making their own LLM these days.
They possibly are (or at least have people doing research), it’s just not very good (yet?) https://aimodels.substack.com/p/apple-is-working-on-multimodal-ai
Remember the early days of Apple Maps?
Remember the early days of Apple Maps?
If that’s an indication, Apple’s AI offerings will someday be as good or better than Google’s. Cause Apple Maps is pretty great these days, but was absolute garbage when they rolled it out.
Apple is working on models, but they seem to be focusing on ones that use tens of gigabytes of RAM, compared to tens of terabytes.
I wouldn’t be surprised Apple ships an “iPhone Pro” with 32GB of RAM dedicated to AI models. You can do a lot of really useful stuff with a model like that… but it can’t compete with GPT4 or Gemini today - and those are moving targets. OpenAI/Google will have even better models (likely using even more RAM) by the time Apple enters this space.
A split system, where some processing happens on device and some in the cloud, could work really well. For example analyse every email/message/call a user has ever sent/received with the local model, but if the user asks how many teeth a crocodile has… you send that one to the cloud.
Tbf, Google has versions of Gemini that will run locally on phones too, and their open source Gemini models run on 16GB of ram or so.
This gives google all the data
So, Apple is lagging on Ai
All speculation, but this particular part of AI, maybe. It’s like claiming Toyota is lagging on cars because they refused to recognize battery electric vehicles as a viable option. No, they do very well with many areas of car manufacturing, but not BEVs
No Big Tech company has your best interests in mind. Ever.
This is the best summary I could come up with:
Apple is looking to team up with Google for a mega-deal to leverage the Gemini AI model for features on iPhone, Bloomberg reported.
This will put Google in a commanding position as the company already has a deal with Apple as the preferred search engine provider on iPhones for the Safari browser.
The publication cited people familiar with the matter saying that Apple is looking to license Google’s AI tech to introduce AI-powered features with iOS updates later this year.
The company’s job listings over last year have suggested that Apple is working on multiple internal and external tools powered by generative AI.
Apple’s own models might power some of the on-device features on the upcoming iOS 18 software update — expected to be announced at the Worldwide Developer Conference (WWDC) historically held in June.
However, the company is exploring partnering with an external provider for generative AI use cases such as image creation and helping users with writing.
The original article contains 372 words, the summary contains 159 words. Saved 57%. I’m a bot and I’m open source!
Both will probably use this as a point in a future anti competitive case against them. “See?! We can totally be friends with each other! We’re definitely not going to use power to benefit only ourselves whenever we get the fucking chance to!”
This should actually work against them. It would be more like “See, we’re not interested in competing, we’d rather maintain monopolies and cartel it up!”
This is a really bad look. It will probably be the case that it will be an opt in feature, and maybe Apple negotiates that Google gives them a model they house on premises and don’t send any data back on, but it’s getting very hard for Apple here to claim privacy and protection (and not that they do a particularly good job of that unless you stop all their telemetry).
If an LLM is gonna be on a phone, it needs to be local. Local is really hard because the models are huge (even with quantization and other tricks). So this seems incredibly unlikely. Then it’s just “who do you trust to sell your data for ads more, Apple or Google?” To which I say neither, and pray Linux phones take off (yes yes I know root an Android and de google it but still).
I don’t see how it’s any different to using Google as the default search engine in Safari.
Also - phones don’t have terabytes of RAM. The idea that a (good) LLM can run on a phone is ridiculous. Yes, you can run small AI models on there - but they’re about as intelligent as an ant… ants can do a lot of useful work, but they’re not on the same level as Gemini or ChatGPT.
It may be no different than using Google as the search engine on safari, assuming I get an opt out. If it’s used for Siri interactions then that gets extremely tricky for one to verify that your interactions aren’t being used to inform adds and or train an LLM. Much harder to opt out vs default search engine there, perhaps.
LLMs do not need terabytes of ram. Heck you can run quantized 7billion param models on 16gb or less (Bloom, Falcon7B — falcon outperforms models with higher memory by the way, so there’s room here for optimization). While not quite as good as openAIs offerings, they’re still quite good. There are Android phones with 24gb of ram so it’s quite possible for Apple to release an iPhone pro with that much, and run it similar to running any large language model on an M1 or M2 Mac. Hell you could probably fit an inference only model in less. Performance wouldn’t be blazing but depending on the task, it could absolutely be sufficient. With Apple MLX and Ferret coming online it’s totally possible that you could, basically today, have a reasonable LLM running on an iPhone 15 Pro. People run OpenHermes 7B for example which uses ~4.4GB to run, without those frameworks. Battery life does take a major hit, but to be honest I’m at a loss for what I need an LLM for on my phone anyways.
Regardless, I want a local LLM or none at all.
Apple partnering with a sub standard LLM / ChatGPT clone. Classic. Tim Cook needs to go.
The article says they’re talking to OpenAI as well. “Exploring” a partnership mean you’re actually going to partner with them - it could just be “what’s your roadmap?”
Apple also “explored” buying Bing and DuckDuckGo.
