whatbrentsay

  • 5/26

    Apple and machine vision

    • apple
    • rumors
    • speculation
  • I've been thinking a lot about much of what Robert Scoble recently theorized is coming in Apple's near future. I highly recommend giving his piece a read, but he sums the fundamental shift up early on:

    ...Apple is about to announce a major shift: from 2D screens, interfaces, and experiences, to true 3D ones...

    Scoble points to 3D maps, a mesh network powered by U-series ultra wideband chips, a machine vision powered Siri, a mixed reality headset that enables a Star Trek holodeck-like experience, spatial audio, and a few other features as evidence. So much of what he wrote about is already in flight to some degree, or rumored so heavily that they're more a question of when than if. A large portion of what he wrote about sounds likely to me.

    A couple things came to my attention recently to convince me that a machine vision powered Siri is coming sooner than later. First, Apple made some incredible strides with iOS 14's VoiceOver feature when it comes to image recognition and descriptions. Seriously, listen to the descriptions in the videos below:

    I had no idea this kind of functionality has been out in the wild for some time on the very phone I use every day. It's beyond impressive.

    The second bit of news was a rumor that Apple will add food tracking to the Health app in iOS 15. The moment I read that rumor, I immediately thought about VoiceOver's subject recognition for the camera and how compelling food tracking would be if it leveraged that technology. Most food tracking apps do not offer a compelling experience because users are tasked with performing detailed and difficult data entry (how reliably can you guess calories based on portion size?). What differentiates the good food tracking apps is that they offer unique features to reduce the friction of manually logging that data. Why would Apple get into food tracking on the back of a laborious data entry experience? Well, I don't think they would; the Health app could just as easily leverage its third party integrations for that. If the Health app is going to do it directly, the experience would have to be better (in Apple's terms) than the status quo. What better way than allowing users to just scan the food in front of them. Considering how Apple likes to test new tech incrementally, this sketchy rumor has a surprising amount of weight to me. Food is a smaller domain (compared to literally everything) and would allow Apple to evaluate the strength of its machine vision at scale.

    Maybe I'm off the deep end but I think there's some smoke here. I expect WWDC to include iPadOS updates that give context to the Pros' M1 upgrade. Now, I'm feeling like a glimpse at Apple's future will actually be present—perhaps not as obviously as Scoble declares—but certainly in some form.