Apple’s AI Woes

For years now I’ve been writing and talking about the trouble with Siri. This problem became even more acute for Apple with the rise of large language models (LLMs) and the world’s collective realization at just how useful a smart artificial intelligence can be.

Last June, it seemed as if Apple finally found religion about making Siri better. At WWDC 2024 they outlined the “Apple Intelligence” strategy that made a lot of sense. While I never expected Apple to build something on par with one of the frontier models, like ChatGPT, I continue to think they don’t need to. If Apple’s AI could remain private and access all my data, that alone makes it more useful than most artificial intelligence. Moreover, as the platform owner, a smart Siri could act as an AI traffic cop, sending more complex requests to the appropriate outside models.

So I think Apple has the right vision, but I’m starting to question their ability to execute on it. Apple has yet to release even a beta of the iOS 18 version with, as one Apple employee explained to me, the “Siri Brain Transplant.” Indeed, Apple recently announced that the advanced Siri features won’t ship in iOS 18 after all. So the brain transplant has been postponed.

Late last year, there was a rumor that Apple is working separately on an LLM-Siri for iOS 19 that will really show how good Siri can be. The fact that there is already a rumor of a new thing when we don’t yet have the improved old thing doesn’t inspire confidence.

It gets worse, though. Mark Gurman, a reliable source, ​now reports the new LLM Siri is also behind​ and its conversational features may not release to consumers until 2027. Ugh. If true, Apple’s failure to deliver on Siri is epic at the Apple Maps and MobileMe launch levels.

The current LLM leaders are evolving weekly. Can you imagine how good they are going to be by 2027? I honestly can’t.

If these rumors are true, Apple is in trouble. It’s not the 1995 Apple-will-they-go-out-of-business-trouble, but it is trouble nonetheless. ​M.G. Siegler suggests​ that if Apple truly is this far behind, they should just default to ChatGPT until they can get their act together. That would be incredibly embarrassing for Apple, but this whole situation is exactly that. It looks like Apple’s AI initiative has a long way to go. Back in the day when the MobilMe launch failed so miserably, people joked that Steve Jobs was walking through the hallways at Cupertino with a flame thrower strapped to his back asking everyone he met, “Do you work on MobileMe?”. When it comes to AI, I think Apple is approaching a flame thrower moment. ​John Gruber agrees​.