discussed the concept of the \"creepy line.\"<\/a> For virtual assistants to work well, and work proactively - anticipating your demands in advance - they need a lot of data on you. At what point does this get weird or \"creepy?\"<\/p>\n\n\t\"While this is a very popular theme for news, I think there's zero creepiness when there'a a lot of utility,\" Marcus said. \"The minute it gets creepy is when a company gets a lot of information and doesn't give anything back.\"<\/p>\n
\n\tWhen I asked Chennapragada, she had a slightly different response - emphasizing user choice. \"I've thought about this problem, oh I don't know, for four years now? There are two fundamental principles. One principle is user control. The user agency. The user needs to be in the driver seat. Take Google Now, for example - basically saying the user is hiring the assistant - they're saying 'hey, use this data for my benefit, here's the set of benefits that I want.' It has to be a user opt-in. That's one.<\/p>\n
\n\t\"There's a complimentary thing about transparency that is, it's not just about agency - it can't be a black box. So in our case, for example, [we say] the location history is useful because when you're on your way to pick up your child from the daycare you kind of need to know where you are, to tell you about the traffic incident ahead.<\/p>\n
\n\t\"These two things go hand-in-hand.\"<\/p>\n
\n\tGoogle Now is just getting started<\/h2>\n
\n\t<\/p>\n
\n\tREUTERS\/Romina Amato<\/p>\n
\n\tGoogle Now isn't where it needs to be just yet.<\/p>\n
\n\tChennapragada didn't discuss usage numbers, but very much frames Google Now as in its infancy. She talks about \"early steps,\" and an elementary focus on\"actually [making] sure the experience works right.\"<\/p>\n
\n\tOver the long term, Google Now wants to be all things to all people. But right now? \"I'll give you the honest picture here ... when we give [users] assistance for enough points during the day or week, [they] find it super useful ... [but] what we do find is in certain markets and certain users ... [we don't have] many useful things to say yet.\"<\/p>\n
\n\tShe highlights a feature in Google Now that helps you remember where you park your car: \"Awesome feature, everyone in the Bay Area who has parking woes, they love it.\" But a user study carried out in Mumbai, India, provided a \"reality check.\"<\/p>\n
\n\tA Google researcher explained the feature's possibilities to a user in Mumbai, \"and the guy just stares at him for a while, and says: 'I jump off a train every day.'<\/p>\n
\n\t\"I'm laughing now, but I was like 'holy sh*t!' We really need to figure out what works for different markets.\"<\/p>\n
\n\tAnd despite being one of Google's chief architects of the virtual assistant revolution, Aparna Chennapragada thinks the state of the tech is over-hyped right now. \"Really, I think we underestimate where we are in the cycle of the whole assistant-based [app], \" she says. \"On demand is all the rage right? ... Assistance is like opinions, everyone has them. The parallel for me is the 1990s web - it's just getting started ... There's an order of magnitude more users, 5 billion people will be online in five years, and suddenly this actually has the ability to affect the real world ... we want to make sure that we're building the blocks right.\"<\/p>\n
\n\tThe five-year horizon<\/h2>\n
\n\t<\/p>\n
\n\tGoogle<\/p>\n
\n\tSo what will<\/em> your phone look like in five years? If Aparna gets her way, it'll be something like this:<\/p>\n \"I want to see two things happen as a user ... one thing is, I really want the technology to be doing a lot more of the heavy lifting. Specifically, for example, I want information and the right information to find me. Today, if you think about these 3.5 billion user [that] have these phones, and they're all turning into broadcasting devices<\/a>. So in a world where everyone's talking, you need selective listening.\"<\/p>\n\n\tIn short: Smart, AI-driven filters, capable of finding the signal in the noise, without the user having to exert themselvdes.<\/p>\n
\n\t\"I think the other thing is: 'What happens to the next generation of apps?' is a really interesting question. I personally think we can disilo and unbundle the function of apps. So apps are a means to an end right? For many of these services they're a way for their functionality to get to users.\"<\/p>\n
\n\tTake Spotify: Users don't care whether it comes via a dedicated app, or a desktop webpage, or an embedded widget in another app, just so long as they get the functionality - music streaming - they want.<\/p>\n
\n\tIt's a vision that is slightly different to Facebook's plans for Messenger, that would see huge numbers of apps replaced by threads between users and companies in the messaging tool, or interactions handled directly by virtual assistant M.<\/p>\n
\n\tChennapragada wouldn't comment on Facebook's plans - but disagreed the future lies in eliminating apps. It's \"less about disintermediating apps [and] more about de-siloing apps.\" Google has made some \"early steps\" with app indexing, letting you search inside apps. But \"even now with Now On Tap, it's so difficult for users to jump between these contexts, jump between these apps. And what we're saying is that needs to be far more fluid ... [It's] less about shrinking them, or obliterating parts of the app, but about saying: 'There's so many more places where parts of your app can surface'.\"<\/p>\n
\n\tHardware limitations<\/h2>\n
\n\t<\/p>\n
\n\tGulfstream Aerospace<\/p>\n
\n\tChennapragada also suggested that the hardware in modern smartphones can still be a limiting factor in what is achievable. She brought up a bug report from a user about Google Now - again, the feature that helps users remember where they park their cars.<\/p>\n
\n\t\"So we got this bug - and I'm not saying this is a normal bug - but we got this bug from a bug report.\" And the report says \"'Hey, I just got off my jet, and you think I parked my car' ... And I was like 'that's not a problem for most of our users!'\"<\/p>\n
\n\tThis highlights the problems Google can face building problems for billions of users<\/a>: There are an unthinkable number of variations an potential use-cases - and right now, \"the hardware signals are not yet there\" to account for all of them.<\/p>\n\n\t\"We could go really invest a lot more in sensors and ten get there, but given battery life, and where things are, it's not really so wise. So we said, you know what, we'll try to make this work with x-percent of accuracy and we live with that.\"<\/p>\n
\n\tSo is the vision for mobile ahead of the reality? \"Not specific to Google Now, but any of these things, when we say this 'contextual computing' as such, there has to be innovation all through the stack.\"<\/p>\n