Apple’s artificial intelligence plan: a thousand little conveniences

Artificial intelligence has become an integral part of every technology company’s discourse for consumers. If you don’t overdo machine learning or neural networks when introducing a new product, chances are you’re looking for crank calculators. This can lead to excessive promise. But judging by its recent performance at WWDC, Apple has taken a smarter, quieter approach.

Throughout Apple’s announcements about iOS, iPadOS, and macOS, there were a number of features and updates that have machine learning at its heart. Some were not announced on stage, and some features that almost certainly use AI were not identified as such, but here is a quick summary of the most prominent mentions we saw:

  • Facial recognition for HomeKit. HomeKit-enabled smart cameras will use photos you’ve tagged on your phone to identify who’s at your door and even advertise them by name.
  • Native sleep tracking for Apple Watch. Use machine learning to classify your movements and detect when you are sleeping. The same mechanism also allows Apple Watch to track new activities like dancing and …
  • Handwashing. Apple Watch not only detects movement but also sound washing your hands, starting a countdown timer to make sure you are washing for as long as necessary.
  • Suggestions from the application library. A folder in the redesigned App Library will use “intelligence on the device” to display the apps you “probably need next”. It is small but potentially useful.
  • Translate application This works completely offline, thanks to machine learning on the device. It detects the languages ​​spoken and can even do live translations of conversations.
  • Sound alerts in iOS 14. This accessibility feature was not mentioned on stage, but it will allow your iPhone to hear things like bells, sirens, barking of dogs or crying babies.
  • IPad Handwriting Recognition. This was not specifically identified as an AI-powered feature, but we bet dollars on donuts. Artificial intelligence is great at image recognition tasks, and identifying Chinese and English characters is a proper challenge.

There are absences on this list, most notably Siri, Apple’s perennially disappointing digital assistant. Although Siri has a lot of AI, this year it received cosmetic updates (oh, and “20x more facts,” whatever that means). A new interface is certainly a welcome change, but it’s a small issue when comparing Siri’s overall performance to other AI assistants.

What these updates do show, however, is Apple’s interest in using machine learning to deliver little conveniences rather than a big unifying “AI” project, as some tech companies with their own digital assistants have promised, which they claim Improve your life smoothly by scheduling your calendar, avoiding your trip, and so on.

This latest project was always going to be a failure, as AI, despite its prowess, is basically extremely good pattern matching software. This has countless uses, and some are incredibly unexpected, but it doesn’t mean that computers can analyze the human complexities of something as common as calendar appointments, a task that relies on numerous unspoken rules about your priorities, routine, tastes, and dislikes. , and more.

The best example of Apple’s approach is the new handwashing feature on the Apple Watch, which uses artificial intelligence to identify when you’re scrubbing the gloves and start a timer. It is a small and silly feature, but it asks little of the user while offering a useful function.

This is a solid tactic for Apple that plays on the company’s long-standing reputation, deserved or not, for delivering software that “just works.” It also avoids the kind of iterative tech-for-tech upgrade that can fail with the average consumer, like Samsung’s Bixby.

Facial recognition on HomeKit cameras is one of the riskiest AI features Apple has announced.
Image: Apple

However, there is also a risk to this approach. Focus too much on convenience, and you may end up overlooking customers’ need for privacy, a mistake Amazon seems to make frequently, like when it started delivering packages inside your house. This could also be a danger to Apple’s machine learning job, despite the company’s continued focus on privacy.

Adding facial recognition to HomeKit cameras, for example, is definitely convenient. The software even connects to the HomePod to announce guests by name as a digital butler. But how will users feel about their photo data being used to identify people through third-party cameras? The videos will be encrypted through Apple’s “Secure Video” framework, but some may still be uncomfortable with the deal. Apple will need to handle this closely, never taking too much control out of the hands of its users, if it wants to continue walking the line between convenience and intrusion.

However, if you can do that, your AI features have a higher chance of slipping discreetly into people’s lives. This, for now, is the sweet spot for machine learning. AI is too dumb to manage your schedule, but it’s smart enough to remind you to wash your hands.