Apple’s AI solution across as an extension of existing functionality rather than a new experience.
As expected, Apple’s introduced a slew of AI features during its WWDC keynote, including a lot of on-device AI and integration with OpenAI’s GPT-4o. In general, most of the new features are things others have introduced before, neatly repackaged by Apple. They looked useful and practical but come across as an extension of existing functionality rather than a new experience.
SVP of Software Engineering Craig Federighi calls Apple Intelligence « AI for the rest of us », echoing early Macintosh marketing. Perhaps the most surprising thing was that it took Apple almost an hour into its Monday keynote to get to AI. But there was a lot to go over, with executives explaining how these features will be integrated into Apple’s operating systems.
Privacy was a major focus. Most of the interactions will be with AI models that run on the device, rather than in the cloud. But to “run more complex requests », Apple will turn to “Private Cloud Compute,” which are servers with Apple silicon « with transparency built-in » (whatever that means). Apple executives said only relevant data goes to the cloud when it has to and it’s never stored. It’s an interesting idea in theory.
Apple is integrating ChatGPT, which can handle larger queries than the on-device models, and has a much broader knowledge about the world in general. Apple said users will have to give ChatGPT permission to run, reminding users that it sends data to the web and might respond with an inaccurate reply (which is true for every LLM application, not to mention to some extent, every search engine).
