Can Apple Intelligence break the pattern of half-baked AI?

First, other companies including Google, Amazon and Microsoft have bragged about similarly useful AI features that have not worked nearly as well as they promised. Second, Apple has a rocky track record in AI, including having foisted on you the dimwitted Siri for more than a decade.

If 2023 was the year when AI mania went into hyperdrive, 2024 must be the “prove it” year.

Too many companies have promised magically useful AI that instead has often been rushed, barely functional, not very useful, insecure or error-prone, including Google’s new AI-powered search that suggested people eat glue. All this half-baked technology is a drain on your time, energy and faith in new forms of AI.

Apple’s new AI-powered features might be great. Most people won’t get their hands on the Apple AI features until September or later, and you might need to buy a new iPhone to use them.

What Apple’s AI demonstration showed (and didn’t)

Apple discussed what are essentially two tiers of AI coming to some newer iPhones, Macs and iPads: the company’s own AI and that of OpenAI’s ChatGPT that will step in for tasks that Siri or Apple’s apps can’t handle.

Apple showed, for example, that if you wanted AI help to write a custom bedtime story for your child, your phone might ask for your permission to share your written prompt with ChatGPT.

It wasn’t clear to me, however, why it wouldn’t be simpler just to use the ChatGPT app or website to generate the bedtime story. (Apple’s own AI may not be capable of this task yet, or Apple didn’t want to risk its AI botching a children’s story.)

Apple also lowered the bar for itself by not promising an all-capable AI assistant that would let you skip the steps in planning your vacation or a grocery order.

No company’s AI can do that, even the AI assistants from Microsoft, Google and OpenAI that say they’re all-capable. (AI can’t even peer into all the places you have relevant information, including a Gmail message, Excel spreadsheet and travel invoice in Expedia.)

Instead, Apple said its AI will make what you already do with its apps — and later, potentially other companies’ apps — smarter and easier.

Several of the features Apple showed — including AI writing helpers and AI for editing objects out of photos — are now standard features from Google, Microsoft or Samsung.

Even with the lower bar, it’s not clear whether Apple can make solid and flawless AI. This is the company that made Siri, the clunky iTunes that even a top executive made fun of, unsuccessful driverless car software and auto-correct that thinks you’re typing “duck.”

The pattern of overpromising and underdelivering

So much AI promises magic and instead is wildly flawed or useful mostly for a small number of tasks.

I don’t blame any of you who aren’t regular users of the new forms of AI — and that’s most people. Only about 7 percent of Americans are using ChatGPT daily, according to a recent survey conducted for the Reuters Institute for the Study of Journalism.

OpenAI announced a chattier version of its chatbot in May, but it didn’t have most of the features the company discussed. Google this year scaled back both its AI image generator and the AI-powered search results after high-profile flubs. Meta AI is kind of dumb. Amazon showed an error-prone AI-upgraded Alexa last year and it hasn’t been released yet. (Amazon founder Jeff Bezos owns The Washington Post.)

Those goofs aren’t Apple’s fault, and the company is usually wary of releasing products before they’re polished.

But the backdrop for Apple’s AI glow-up is that it and other companies are under so much pressure to show off their AI expertise to investors, employees and business partners. It’s hard to know if any AI is really made for you, or if it’s mostly for corporate self-interest.