Who texts with autocorrect when they can speech to text? If hands are too busy to type, talk. Re-engineering the problem's often easier.
LLMs that are changing the world only fit well in Apple's polished world when they reach the quality needed to not mess anything up. There's a truthyness to Apple's world, a verified truth, a positivism (the epistemological version of the world) that LLMs by-and-large strive for in a very different way from the intelligent design of Apple.
The most immediate way LLMs will change the world is by making fusion reactors viable through market forces as they take the baton from blockchains in burning power.
The best chips in every device capable of running LLMs today and they will only get better. When all of compute is AI, users will not want to put their entire lives into a closed company's AI. Apple will be perfectly positioned to let users own their AI, so the AI doesn't own them. Their strong stance on privacy makes this even more realistic.
They control the OS. As the use of apps die and all of compute becomes AI, they will have the tightest integration of AI in the world, if they make the right moves allowing them to ship fully integrated and optimized AI only machines. That means stripping out everything not needed for AI and focusing on functional capabilities exposed for use by AI.
If they are willing to bite the bullet when they see that apps will be needed less and less, they can focus all of the brilliant team they have on a much more simplified model and eliminate the complexity of interaction faced with Siri and the old model of AI.
Apple was operating with "old" AI and can now integrate far superior AI to make it perform far better and provide a superior user experience.
The biggest problem Apple faces is realizing they will lose significant revenue as the integration of AI, agents, and functions makes them unnecessary and we'll have to see if they can bite the bullet and radically reshape the company around it. Their hardware and control of the OS and functions makes them in a far superior position to any company in tech to win. They can't rush it though. They have to do it right and have a plan for the migration to what is inevitable.
Already seeing them working hard to do the research needed in the AI sphere. MLX performs beautifully. Their experimentation in small LLMs is brilliant, because they'll be able to deploy many small llms and agents with specialized functions in the short term, that can run on previous iPhone versions while they upgrade the hardware in future generations for llms everywhere.
They have some of the smartest people in the world on their team and they have the money to recruit more if they can share a vision of domination. I think it is possible, I just am not privy to how they are thinking about the future. If they try to hold on to their current model too long they could lose.
(2) Microsoft. When it comes to mobile, crickets.
(3) Meta. Can execute pretty well on VR and other things but it doesn't matter when your main product is so hated. The big picture is that "FAANG" companies comfortably dominate certain domains and find it almost impossible to innovate because nothing they come up will get the same monopoly profits their business is built around.
(4) Why not let others take the risk and take a 30% cut off the top at the app store? It's a lot less embarrassing than Apple Vision Pro.
(5) Apple is doing really well in terms of hardware support for NNs. Microsoft and Intel are talking about AI accelerators, Apple has them in all their devices. Because of the UMA, Apple's M* can run bigger LLMs than all but the absolutely largest NVIDIA cards, even if they don't manage the same GFLOPS.
(6) Everybody struggled with voice assistants. Alexa, Siri and Cortana were all disappointments.