Apple Intelligence's biggest problem isn't the Intelligence-it's Apple

The key to Al success is to make the features a pleasure to use. That's not happening right now.

- BY JASON SNELL

Everyone knows that Apple is playing catch-up when it comes to Apple Intelligence. The company's shipping Al models seem to be way behind the cutting edge, as OpenAl grows, Google pushes forward, and newcomers hit the scene.

I'm sure Apple is pouring everything it can into building better, more modern models, and we'll hear about that effort in detail in June at the Worldwide Developers Conference. But what troubles me most about the Apple Intelligence rollout isn't that Apple was caught flatfooted by the Al hype train and is struggling to catch up—it’s that Apple’s implementation of Al features also feels slapdash and rushed.

Apple doesn’t have to end up with the best large language model around to win the Al wars. It can be in the ballpark of the best or partner with the leaders to get what it needs. But it can’t fail at the part that is uniquely Apple: making those features a pleasure to use, in the way we all expect from Apple. Right now, that’s where Apple is failing.

APPLE’S BEST SHOT AT AI’S WORST

The worst thing about Al is that since much of it springs from the concept of a text-based language model, Al interfaces tend to be empty text boxes that you have to type something into. I can’t believe we're back here. This is serious pre-1984 thinking, 40 years after Apple put a stake in the heart of the command-line interface.

Giving users an empty text box and expecting them to know what to say to get the result they want is a colossal user-interface failure. An empty text box is cruel. (And no, having to carefully issue abstract commands via voice is not a good alternative, nor is forcing users to laboriously correct mistaken output with more text entry.)