Future Tense

Google’s New Messaging App Is “Smart.” But Should You Use It?

Google’s new messaging app asks us to trade privacy for artificial intelligence.


The first thing to remember about Allo, Google’s new messaging app, is that there are a lot of messaging apps out there. For that matter, there are a lot of Google messaging apps out there, including Hangouts and Duo. (As my former colleague Lily Hay Newman pointed out, there really should be just one—and it should be called Gchat.) The tech press may be abuzz over Allo’s launch, and there are good reasons to pay attention to it. Whether there are good reasons to use it, however, is less clear.

As a messaging app, Allo seems … fine? Pretty good, really, to the extent that the major messaging apps are distinguishable from one another. If your goal is to communicate with someone via text or voice recording or funny animal sticker, it will do the trick. Clean interface, simple controls—it’s straightforward and pleasant to use, which probably means the teens will hate it.

What’s special about Allo is not that it’s supposed to be a better messaging app than all the others, per se. Rather, Google is billing it as a smarter messaging app. “Smarter” refers to the artificial intelligence built into the Google Assistant, a sort of helper bot that lives in the app. You can converse with it directly, like Siri. But it can also slide into your conversations with others by suggesting replies to their messages, offering a restaurant recommendation, or connecting you to relevant info from the Web or other apps. It’s like a genial butler who is also sort of nosy.

This is not a novel idea: Bots are de rigeur in messaging apps these days. But Google has a big head start on rival companies when it comes to both A.I. and search. That comes in handy in figuring out what you mean when you type, for example, “who won the game last night,” or “when’s my next appointment.” Apple’s Siri has long been criticized as useless or unintelligent because it seems to struggle to even hear you correctly, let alone answer basic inquiries. It may be improving but Google’s voice and conversational search technology has long felt superior. “Smart” may still be the wrong word for the Google Assistant, at least by the standards of human intelligence, but “less dumb” is a step in the right direction.

Part of the promise of Allo, then, is that of a messaging app that is not simply a line of communication between two or more people, but a portal through which they can readily draw in information from the outside world. It’s the kind of thing that sounds impressive in a Silicon Valley boardroom or in the pages of a tech blog. It’s an open question whether these sorts of features really matter to those who just want a convenient way to chat with friends or family.

More entertaining is the “smart reply” feature that suggests potential responses to people’s messages. You can have some fun or laughs at its expense by seeing how it responds to a booty call, or a Welsh corgi. But the real utility lies in the convenience of being able to quickly tap on a canned response, rather than typing something from scratch, in the many situations when nothing more elaborate is needed. Google said earlier this year that 10 percent of replies in its email app Inbox begin with a “smart reply,” which users can either send as-is or tweak as needed. As Google gets better at this, the number should go up. And it seems even more appropriate to a chat app than the somewhat more formal realm of email.

Allo’s A.I. assistant suggests “smart replies” to your friend’s messages.

Image courtesy of Google

A clever-sounding twist is that the Google Assistant will actually pay attention to the language you use in chats with a given person, so that it can tailor its smart replies to the context. If you’re in the habit of saying “yo” to your friend but not to your grandfather, it will adapt accordingly.

This is all quite neat and futuristic, if not exactly essential, from the user’s perspective. But, as we’ve come to expect from Google’s services, there’s a flipside for those concerned with privacy.

Google raised the hopes and expectations of privacy advocates when it announced in March that Allo would offer end-to-end encryption, and that it would delete your messages after a certain amount of time rather than storing them indefinitely. In a surprising reversal, Google disclosed when it launched the app on Wednesday that it actually will store your chats indefinitely by default, until or unless you actively delete them. If you want your chats to be truly private, you’ll have to switch to “Incognito” mode. This will disable the A.I. assistant features that were supposed to be Allo’s selling point. Google told the Verge it made the change because the smart replies wouldn’t work as well without access to people’s chat histories.

There’s nothing inherently evil about a messaging app that stores your chats or makes end-to-end encryption an option, rather than the default. And users of Google services like Gmail should be familiar with the tradeoffs involved in letting the company store and scan your personal communications. But Google’s backtracking was a big strategic blunder, because the big story of the day of Allo’s launch was not its “intelligence” but its privacy risks. Edward Snowden spent half the day on Twitter telling people not to download it.

Eventually the privacy backlash will fade, but it underscores what might be a bigger problem for Google and other big tech companies. As I explained at length in a story about the rise of A.I. assistants, Silicon Valley’s dream of making them your primary portal to the online world is partly about making things easier on users. But it’s also about ingratiating tech companies’ own services more deeply into our lives, so that we come to trust and depend on them—and give them ever more access to our data. When we let a chat bot in on all our daily interactions, we’re letting its creator in on them too.

Allo is a test case for our willingness to participate in that bargain. Based on the press it has received so far, it’s probably a test Google wishes it could retake.

Previously in Slate: