Microsoft's new Clutter feature for cleaning up your inbox automatically is the latest of several tools based on machine learning that the Office team thinks are ready for prime time. In fact, it's the latest of many machine learning tools that researchers at Microsoft have developed over the years. But this time, Microsoft is confident enough to make them broadly available.
Like Delve, which suggests relevant documents and discussions from your colleagues, it uses the Office Graph to understand who you actually connect to and collaborate with and who is just interrupting you. Delve makes lists like "presentations you've seen in meetings" by cross-referencing who was at the meetings you were invited to, what files they've saved and the timestamp on files that were presented at the right time to be in the meeting.
That could feel creepy if you think about the system watching you, and it's certainly confidential information you wouldn't want your competitors to see, so it won't find files that haven't been shared with you. But when you use it, if Delve gets it right, it just feels useful.
Clutter will take messages that aren't spam but aren't actually useful either, whether they're special offers from a loyalty card or too-widely distributed messages about events happening in a distant office. Clutter uses the Office Graph to know who you're connected to as well as watching how you've reacted to similar messages.
Machine learning makes it more sophisticated than filing messages from iTunes just because a lot of other people have filed those messages; if you always open and read them and click the links, Clutter isn't likely to hide them away.
This idea of automatically triaging and arranging email isn't new. Google's new Inbox tool has some similar ideas -- but so did Xobni, a startup that Yahoo bought some years ago that tried to turn your inbox into a social network by showing who you interacted with most often.
Outlook.com, Microsoft's consumer webmail service, has offered to sweep newsletters into their own folder automatically for quite some time, but it hasn't moved personal emails. And back in 2006, Microsoft Research produced a tool called SNARF that let you sort your inbox to prioritize messages from people you frequently mailed and replied to.
But until this year, even though machine learning has advanced significantly, Microsoft included only simpler tools from its research group in Outlook. The warning if you press Send on a button that mentions an attachment without attaching a file, and the variably successful Suggested Meetings feature that extracts times and places from incoming email, ready to turn into an appointment, both started as research projects.
Sign up for Computerworld eNewsletters.