For the multiple times Apple executives mentioned “machine learning” at the company’s developers conference Monday, they also emphasized an older theme that could be more important: Unified software.
Apple’s mobile operating system, or iOS, works the same and runs the same on every iPhone. That’s why app developers often prefer building for Apple before they build for Google’s Android, an operating system that’s splintered across different kinds of hardware. While the latest versions of Android consistently run on higher-end phones like the Google Pixel, cheaper and older phones often only run previous versions.
Apple’s control over both its operating system and the hardware on which that software runs came up throughout the keynote. That highlights what Google really needs to worry about when competing with Apple: The fragmentation of Android. This matters even when Google has a technical advantage.
Apple’s latest announcements in augmented reality are where the company’s advantage with a unified platform showed the most.
The ability to quickly deploy software to devices immediately differentiates Apple’s AR from Google AR computing platform Tango, which only runs on select newer-model phones.
“When you bring the software together with these devices, we actually have hundreds of millions of iPhones and iPads that are going to be capable of AR,” Apple’s head of software, Craig Federighi, said. “That’s going to make overnight the AR kit the largest AR platform in the world.”
In general, only a portion of Android phones tend to run the latest version of the mobile software at any given time. While Google builds its own hardware now — such as the Pixel phone — it still relies largely on third-party manufacturers and carriers to deliver Android updates to hardware that runs it.
Update: Google said it has made efforts to lessen the fragmentation of software across Android devices, including rolling out a program to make it easier for device makers to upgrade to newer versions of Android, and ensuring software in Google Play Services is updated frequently across devices. Google reports 93 percent of users ahave the latest version of Google Play Services.
Federighi brought up hardware and platform advantages again when he announced a new set of machine learning software tools for developers.
Developers that used Apple’s new machine learning tools would be able to execute their use with “tremendous performance on-device,” he said, and have access to “all the data privacy benefits and all of the carefully tuned compatibility with all of our platforms.”
What he didn’t seem to emphasize was what made the tools themselves interesting. That’s probably because Apple is known to lag behind Google, Microsoft and others in machine learning.
The iPhone didn’t publish its first artificial intelligence research paper until December. By comparison, Google had 44 papers accepted this year to the International Conference on Machine Learning while Microsoft had 33, according to a Medium post by a researcher at AI think tank OpenAI.
And Apple’s new machine learning offerings are also not all drawn from Apple technology. “Some of the pre-trained machine learning models that Apple offers are open-sourced Google code, primarily for image recognition,” noted Quartz reporter Dave Gershgorn. Google confirmed this.