Gmail AI: What's Privacy concerns about Smart Reply?

Posted by John Onwuegbu | November 09, 2015 |

Google's cutting-edge machine learning intelligence introduced with Inbox by Gmail app, is perhaps the best message management enhancement ever to hit the email system. The Artificial Intelligence (AI) powered app is able to automatically sort out your priority emails and also suggest/set reminders on calender events.

Now, the company has unveiled additional feature, Smart Reply, which enables the Inbox by Gmail app to analyze email contents and auto suggest appropriate responses.

As always, skeptics have already started to question the level of users' privacy exposed in the background processing involved in Gmail Smart Reply.

Gmail Smart Reply is built on a system of "recurrent neural networks", with "long short-term-memory" (most useful in predicting an accurate response) devoid of any human interception, meaning that no human is reading your email messages.

The machine learning works on data set that the developers' themselves does not have access to read and thus eliminating any privacy concerns.

How the technology works?

According to Google, the encoding of email is handled by one network, which after consuming the words of a message, produces a list known as vector. The vector captures the meaning behind the message, then a second network utilizes "the vector" to create an appropriate reply.

The feature has been rolled out on Inbox by Gmail for Android and iOS, so if you already have the app installed, you'll need to update to get the new AI experience.