The next semi-annual update to Windows 10 will use machine learning models to make automatic rebooting for updates a bit less annoying. The models will attempt to predict when you’re likely to return to your PC and not update if you’re expected back soon.
In prior versions of Windows, it was routine for systems to be compromised through flaws that were patched months previously because Windows users deferred installing those updates or even disabled Windows Update entirely. Windows 10 goes to some lengths to ensure that Windows users, especially home users, apply the monthly security patches in a timely fashion through a policy of automatically rebooting when a patch is available. Last year, Microsoft gave users greater control over this feature, allowing those reboots to be explicitly scheduled, but the policy of automatic installation and rebooting remains fundamentally in place.
Currently, Windows will detect if you’re away from your system (mouse and keyboard idle and not playing video or anything comparable) and perform its reboots during those idle moments. However, at the moment, the system doesn’t distinguish between briefly stepping away from the machine to grab a cup of coffee and being away for hours because you’ve left the office or gone to bed. This has provoked some amount of complaining due to the updates interrupting work.
With the new predictive system, Windows will try to distinguish between these two cases, and it will avoid the update if the absence is expected to be short. This in turn should remedy the situation of returning to your computer, coffee mug in hand, only to find it in the middle of rebooting. Microsoft says that the model has proven effective in internal testing, and it will undergo further training and updating based on user feedback.