Software updates are a problem. Both software-makers and hardware-makers want you to keep your devices up to date; updates contain patches and bug fixes that ensure your device runs as quickly and securely as possible, in addition to delivering new features. But the act of updating a device is never convenient. Depending on the machine and operating system, users may opt to delay updates ad infinitum rather than wait the two to 20 minutes for their systems to reboot. To combat that possibility, Microsoft has taken a more forceful approach, automatically checking for updates and installing them as needed. For some, this has proven frustrating—the device may decide to update itself when you’ve only stepped away for a moment to get coffee. It may even decide to reboot when you’re in the middle of a task.
Rather than a snooze update button (which has failed to solve the issue for some), Microsoft is taking a new approach: using machine learning.
“Have you ever had to stop what you were doing, or wait for your computer to boot up because the device updated at the wrong time?” Microsoft’s Windows Insider chief Dona Sarkar writes in a blog post. “We heard you, and to alleviate this pain, if you have an update pending we’ve updated our reboot logic to use a new system that is more adaptive and proactive.”
For the new update system, Microsoft trained a predictive model to more intelligently decide when is a good time to reboot your PC to install an update. The system will check if the device is currently in use, and when not in use, it will try to predict whether you’ve just left the computer for a short break or for a longer spell.
Using artificial intelligence and machine learning to decide when to update a computer seems like overkill, but it is a legitimate, problematic issue no matter what device you’re on. On iOS, users aren’t forced to update their device in the same way that Windows 10 users are. However, Apple does a good job of guilting users into updating their devices by placing a nagging red update icon in the upper right of its Settings app, although users can avoid this by switching on automatic updates. The technique works pretty well: As of June, 81 percent of iOS users are on iOS 11 (although fewer are likely running the very latest version, iOS 11.4.1).
On desktop, Apple issues updates periodically—something you typically learn thanks to an “Updates Available” popup in the upper-right corner of the screen. If you don’t have Auto Update switched on, macOS gives you the option to restart right then, in an hour, that night, or to “remind me tomorrow.” With the latter, Mac users are able to put off sometimes important system updates for days, weeks, or months until they finally decide to accept the update. The latest versions of Android take a similar approach to Microsoft, updating your device automatically over the air, in the background. Automatic updates are the default on Android, where they’re opt in on iOS.
Microsoft’s decision to use machine learning to try to solve the problem of software updates highlights the struggle between software-makers and users who are lax about keeping their devices’ software up to date. Threats such as January’s Spectre and Meltdown discoveries make it clear that keeping devices updated can be vitally important to securing consumers against data theft and hacking. Both automatic software updates and manual updates have their problems: The former may happen at inopportune times, while the latter means never-updaters could leave systems unprotected against the latest digital threats. Microsoft’s new machine-learning powered updates are currently available to beta testers and, fittingly, should roll out more broadly to Windows 10 users in future software updates.