clock menu more-arrow no yes

Filed under:

Every keystroke a Google Search™

New, 19 comments

Google's Allo messenger turns your conversations into a product

I like using Google to search the internet. Sometimes I'm looking for websites, and Google is a great place to find them. Of course, Google's only product isn't Google Search. They also do other things! Like, did you know Google will store a bunch of emails for you and show them to you whenever you want? They also make a whole operating system for phones called "Android." The company really is a jack of all trades. What will they think of next?

MAYBE IT WILL BE A WAY TO READ EVERY SINGLE CHARACTER I TYPE INTO ANY KEYBOARD ANYWHERE AND USE MY INPUT TO SEARCH AND REFINE ITS KNOWLEDGE GRAPH IN A BID FOR ULTIMATE WORLD DOMINATION OR AT LEAST THE ROLE OF HELPFUL, FRIENDLY AI ASSISTANT WHO KNOWS YOU BETTER THAN YOU KNOW YOURSELF, LIKE A GOOD FRIEND OR LOVER.

Last week it was Gboard, and I was all like, "Hey, this is pretty cool! I can search for GIFs and restaurants right from my keyboard without leaving the app. Google's autocomplete suggestions are also pretty great. Huh, who would've thought."

This week Google announced its Allo messenger app, which looks at every message you receive and offers suggestions for what you could say. This suggested reply feature is already in Google Inbox (that cool email service I was telling you about) and I always ignore it. But the Allo incarnation is even creepier. It's trying to talk like me, and it can even identify what's in a picture and suggest a response:

THIS SOUNDS LIKE THE DEATH OF LANGUAGE.

It's like Facebook's emoji reactions, but inserted into the most intimate communication medium we have outside of talking out loud to each other. Every emotion expressed by multiple choice questionnaire. No time to figure out how I feel about your dog, I'll just use a canned reply. We're becoming the very thing we hate: automated phone systems. "I'm sorry, all our operators are busy right now, but on behalf of Paul Miller I would like you to know that Google's AI loves your dog."

And here's the scary part: Google's only going to get way, way better at this. Now that it can hover over the shoulder of every single Allo conversation, it's going to get so smart. It will be so good at judging dogs. Google learns from how you talk, and it really truly wants to predict how you "would" respond. It's going to be truly impressive.

But what if you're the sort of person who "would" just pick a canned reply offered by Google's AI? What then? That's a dangerous feedback loop. It's obvious that Google will shape our language by trying to predict it. Google has to be proactive and selective, or it will end up with a Microsoft Tay situation: corporate robots that say racist things because that's how they were "trained." This is one reason some humans throughout history have decided to say their own words, because having someone else speak for you can be sub-optimal.

Allo is, really and truly, one of the most interesting new interfaces for AI. An AI that doesn't just speak when spoken to, but is engaged in every syllable and selfie of conversation. It's the perfect recipe to get way smarter, and to give people way more of what they want in an AI assistant. An AI that really knows you can be better at working with your schedule, filtering spam and abuse on your behalf, and will probably even knows what I want for dinner before my stomach does.

But I won't let it tell me how I feel about your dog.