Google’s context auto-replies draw mixed reviews
NEW YORK — Google is toeing the line between helping you save time and creeping you out as it turns to machines to suggest email replies on your behalf.
The customized autoresponses come in the latest version of Gmail on the web and expand on a feature already available on Android devices and iPhones. They’re just one more example of how artificial intelligence is seeping into everyday online life, whether it’s to tailor product recommendations or correct spelling.
So far the new feature has been drawing mixed responses from users.
The new feature, called Smart Reply, offers three short responses, like “It was great seeing you too,” or “I’ll look into it.” Unlike standard auto-replies when on vacation, for instance, these are customized to an individual email based on its context. If you select one, you can either send it immediately or edit it before sending.
The responses are automatically created using Google’s artificial intelligence systems. Humans aren’t reading people’s emails, but machines are scanning them. Although Google stopped scanning email to target advertising in 2017, it still scans them to filter out junk mail, identify phishing scams and, now, to create suggested replies. (Yahoo and AOL, both owned by Verizon, still scan email for advertising.)
Google’s suggestions draw on what’s in your email, such as the text of the message and the subject line. The analysis can include past conversations. For example, if someone says “Thanks!” more often than “Thanks,” with no exclamation point, the suggested response would likely reflect that.
It’s not known, though, how far Google goes in scanning your email. For instance, Gmail scans attachments for security risks, but is it also using that information to customize responses? The machines behind Google Photos can tell whether an image has food, flowers or the beach. Might Gmail suggest, “How was lunch?” Google didn’t respond to a query for details on what it scans for replies.