Google Developing Deep Learning AI to Reply to Your Millions of E-mails for You
Just ... don't judge me or the way I e-mail, okay?
Google’s cooking up some neat stuff for their next Gmail innovation: Smart Reply, an AI that will analyze incoming e-mails and provide you with three options to reply back with. On the surface, it seems (kind of) simple, but as the WIRED article describes, there’s actually a Hell of a lot more at play here than just offering you canned messages to fire off.
For starters, it’s not actually a single AI that’s at work here. Smart Reply consists of a complex system of deep learning AIs, each with a very specific task. One AI analyzes the incoming e-mails to find out what’s being said. The other then takes that data and tries to figure out possible responses—one word at a time, much like how you’d normally write an e-mail.
Frankly, the idea of unleashing deep learning AI on all our e-mails is at once awe-inspiring and frightening. Call me a luddite, but it’s already worrying that so much of our lives are tied in to just a few major internet “networks” (think Facebook, Twitter, or Google), so to allow one of those use an AI like this to essentially learn and recreate the way we speak to each other seems … odd.
The implications of technology like this are stunning, in both good and bad ways. I’m excited to see what happens, for sure, but as Stephen Hawking and many other tech luminaries suggest, we need to ensure that we maintain the “human aspect” of tech development.
This is not to say that this is the line drawn in the sand—after all, Smart Reply is still being developed and has its fair share of kinks. Perhaps the most spookiest thing about it is that it consistently tries to add in “I love you” as a reply option. Heh … how cool would it be if that was the AI trying to talk to us?
—Please make note of The Mary Sue’s general comment policy.—