Apple Intelligence is boring and that might be

Apple announced Apple Intelligence at WWDC and the reaction from the AI community was a collective shrug. “That’s it? Better autocomplete? Smarter Siri? Notification summaries?”

Yeah. That’s it.

And I think it might be the smartest AI strategy of 2024.

What Apple didn’t do

Apple didn’t build a chatbot to compete with ChatGPT. Didn’t launch an art generator. Didn’t announce a reasoning model. Didn’t claim they were pursuing AGI. Didn’t release a paper claiming state-of-the-art on any benchmark.

They didn’t play the game everyone else is playing.

Instead, they took a bunch of models (some on-device, some in the cloud) and used them to make existing phone features slightly better. Writing suggestions in Mail. Photo search that understands natural language. Notification prioritization. A Siri that can actually do things across apps.

None of this is exciting in isolation. All of it, together, on a device that two billion people carry every day, might be the largest deployment of AI in history. Not by capability. By reach.

The Apple bet

Apple is betting that most people don’t want a chatbot. Most people don’t want to generate art. Most people don’t want to reason about complex problems with an AI assistant.

Most people want their phone to work better.

Find the photo I took of that receipt at the restaurant in Portland last March. Summarize these 47 notification I missed during my meeting. Write a reply to this email that sounds like me. These are the mundane, everyday friction points that AI can solve without anyone needing to understand prompts, context windows, or model architectures.

Apple is integrating AI the way they integrated the internet in the early 2000s. Not as a product. As a feature. Not something you go to. Something that’s there, underneath, making everything else work better.

Why boring might win

The AI industry is obsessed with capability. Who has the best model? Who scores highest on MMLU? Who can solve the hardest math problems? Those are real metrics and they matter for certain applications.

But for the 3 billion people who use smartphones every day, capability beyond a certain threshold is irrelevant. Can it understand what I’m looking for? Can it write a passable email? Can it sort my photos intelligently? If yes, the model doesn’t need to be the best. It needs to be good enough, fast, and integrated.

Apple has the integration advantage that no other AI company has. They control the hardware, the operating system, the apps, and now the AI stack. They can run small models on-device for privacy-sensitive tasks and route complex queries to the cloud. They can embed AI into every pixel of the interface without requiring the user to open a separate app or learn a new interface.

That’s boring. That’s also how technology actually gets adopted at scale. Not through demos and research papers. Through making the thing you already do slightly less annoying.

I think in five years we’ll look back at 2024 and see two AI strategies. One was the AGI race, the pursuit of the most powerful model, the most capable reasoning system. The other was Apple’s strategy: make the phone better.

Both strategies matter. But the one that affects more people’s daily lives is the one that ships on every iPhone.

And that’s the boring one.

Related thinking: