Using Artificial Intelligence to make Technology disappear — Rand Hindi @ #ai2015
Rand Hindi recently spoke at the Playfair Capital Machine Intelligence 2015 showcase about how we can use artificial intelligence to make technology disappear.
Dr Rand Hindi is an entrepreneur and data scientist. He is the founder and CEO of Snips, an Artificial Intelligence company specialized in context-awareness. Rand started coding at the age of 10, founded his first startup at 14 and created a web development agency at 15 before starting his PhD at the age of 21. He has been elected as a TR35 by the MIT Technology Review, as a “30 under 30” by Forbes, received the Excellence Française award, was the founding ambassador of the Sandbox network in Paris, is a World Economic Forum Global Shaper and a Kairos Society fellow. He holds a BSc in Computer Science and a PhD in Bioinformatics from University College London (UCL), as well as two graduate degrees from Singularity University in Silicon Valley and THNK in Amsterdam.
Rand began by describing an all too familiar experience that we can all relate to from our day to day lives. “About a month and a half ago, I went on holiday with my girlfriend to Costa Rica. If you’ve never been there, it’s actually pretty awesome. It’s beaches, jungles, but most importantly, there is no connectivity whatsoever. What that means is that you essentially spend your day just being uninterrupted and interacting with the real world as it used to be a few years ago. Unfortunately, what happened is every time when we got back to the hotel every evening, all of a sudden, we got connected to Wi-Fi and we started getting bombarded by push notifications and people calling us and emails that we needed to reply urgently. We got interrupted all the time. We got interrupted when having a dinner date, when we were just sipping a Mojito, when we were actually even making love. It just kept on buzzing and ringing all the time. That would be okay, right? You could just put your phone away. The problem is we also kept on checking our phones all the time. Even when we did not receive a notification, we would just take a look, open the phone, just in case we had received this one little notification that’s important that day. It could be a message from someone you actually love. It could be Ashton Kutcher following you on Twitter, something that you think is worth checking your phone for every 5 minutes.”
He describes a very real and unhealthy addiction that people have with technology today. An addiction that has become so profound that 9 out of 10 people experience something called phantom vibrations. Rand says “This is when you think your phone vibrated in your pocket, but it didn’t. How often in our lifetime have we been so deeply addicted to something that our brain started imagining hallucinating all day long?”
The reactions we have developed to these constant notifications are similar to the Pavlov’s dog concept whereby a potent biological stimulus becomes expressed in response to a previously neutral stimulus that elicits the desired response — or more commonly known as classical conditioning. “In this case here the bell is the notification and the sugar is this one tiny piece of information you care about.”
The chart above shows the history of connected devices together with how they are expected to increase over the coming years. “When you look at the history of connected devices back in the 1990s, you had something called the unplugged era — basically no connected device. The internet then arrived in 1995 and you had your first connected device, your computer, which started sending you emails. Some of them were spam. Some of them were important stuff, but essentially you had your first form of intrusiveness of technology in your daily workflow. Then comes the mobile era in 2005. it’s not one device but 3 of them. A phone, a computer, a tablet. The problem is that each of these devices have no idea about each other. Therefore, the strategy is to interrupt everywhere all the time. If you have, for example, a Mac and an iPhone, if someone calls you, it rings on both of them. Even when you reply, it keeps ringing. How annoying is that?!Try to imagine what’s going to happen with the Internet of Things. We’re expected to have 100 billion connected devices by 2025. That’s 14 for every person on the planet. Just for a second, try to imagine yourself with 14 devices, none of them being able to synchronize, just basically pushing notifications all day long. That’s not including things like your light bulbs, which are going to be connected as well. You could say that the number of connected devices is proportional to the amount of friction technology brings in your daily life. If you keep looking at the trend, it’s actually exponential, so by 2030, it’s going to be maybe 1,000 devices. I don’t know about you guys, but I definitely don’t want to be living in a world like that. Fortunately, the cool thing about AI is that it can be as a type of anti-friction. We’re not talking here about Artificial General Intelligence (AGI). We’re talking about a specific domain, which is called context awareness.”
The domain of context awareness involves giving devices the ability to sense and react to the context of the user and interact accordingly. For example, your push notifications could be filtered automatically based on what you’re currently doing. If you’re in a meeting, there is no point sending you social notifications.
Rand explains, “When you look at the growth in the trend of AI, what you realize is that it takes a little bit longer to start, but we did hit that inflection point now. When it actually starts growing, it goes much faster. If at first you start as exponential, this could soon develop to become double exponential. What that means is that if AI is anti-friction, at some point when it becomes capable enough, it will actually be more capable than the problem that we get with connected devices. At that specific point, you’re going to reverse completely the interaction we have with devices. All of a sudden, it’s going to drop down to zero. This is what we call ubiquitous computing. It’s essentially the idea that you could adding devices. It does not add friction. It actually adds value. You could have a million devices at that point. It keeps adding value. This is important because when this becomes really ubiquitous, once you get used to the idea that all these devices are working for you in the background and not interrupting you, then you’re going to stop paying attention.
If you stop paying attention, it will disappear from your consciousness, and the feeling will be that the world is unplugged with all of the power of technology working for you in the background. It will feel like the 1990s, but with all the technology today. I think this is pretty cool because what it means is that for the next 5 years, 10 years, it’s going to be horrible. I’m not going to lie to you. It’s going to get really bad, like really bad, but down the road, it’s actually pretty cool. We just have to hold just enough time.”
Rand offers an insightful view into how our lives will change through technology.
So where does Snips come into this?
Snips are building a context aware mobile application, which analyzes your habits to enable a faster, more ubiquitous and seamless access to each of your services and applications on your phone or your wider collection of connected devices.
Rand makes an analogy of artificial intelligence to electricity. “When you think about electricity back in the 1800s, it was very expensive to produce. It was unreliable, would cut all the time. It was very dangerous. Your apartment would catch fire. People actually believed at the time that oil lamps were safer. As it became better and better and more mature, we stopped thinking about it and today you probably don’t realize, but you have electricity so deeply into your life. It’s around you. it’s in your pocket. it’s in your heart for your pacemaker, but you don’t even think about it any more.This is basically what we’re trying to do here with technology.”
Snips is building mobile based context awareness through 4 different layers:
The first layer comprises of a social layer that is attempting to model the way that people communicate with one another and predict who they will want to speak with next. An example of a way to do that would be to look at your emails and try to figure out in which context are you sending emails to whom and who is actually inside the same thread as the other person. This essentially gives you the ability to recreate a kind of relationship between people.
Rand explains “Now that you have this kind of information, you can also use it to store disambiguating things like calendar events. You may for instance have an event in your calendar called, “Meeting with Michael,” but which Michael are you referring to? This is important because if you want to have any kind of personal assistant, you need to be able to infer that information. Then you can figure out that in the context of that meeting at that time at that place, it’s more likely to be Michael X than one of the other Michael’s I know.”
The second layer is about figuring out what you’re currently engaged in in your real life. Snips believe the best way to do this is to monitor your location data. “Location data is essentially a trace of what you’ve been doing throughout the day. If from it, you can extract the specific places you’ve been at and re-contextualize that, then you can say, “Right now, you’re at a conference on AI. Therefore, what you might be interested in is maybe the profiles of the people who are talking or maybe you just want your camera to record or something like that.”
The third layer is more of an aggregation layer that connects your social layer, location layer with all the other relevant applications on your phone. “You can start essentially creating really accurate maps of what’s happening on a population level. Here, for example, in Paris, we did a prequel experiment where we aggregated location data for 200,000 people to measure very precisely the population flow in public transport. From this, you can start predicting how many people will be onboard the trains, which is actually pretty nice because then now you can adjust the schedules of the trains accordingly, as well as tell people that, “Perhaps you should take one a big later if you want to be comfortable. The same kind of idea could be used to model things like the risk of having a car accident.”
In another experiment, Snips tried to figure out what is causing car accidents in London. They modelled street topology and weather. Without any prior knowledge about a specific place, they were able to predict that places like Trafalgar Square are indeed a lot more dangerous. This in turn can be put back into a self-driving car, who now all of a sudden have an awareness of the context of risk it’s exposed to.
You could start telling it, “I’d like you to go and pick up my kids at school by minimizing the risk of having an accident on their way back home.” This is the promise of technology is that you want to be able to have all of these super intelligent agents all over the place working together to provide something that is really, really, really non intrusive and powerful.
“When you put all layers together: social layer, individual personal layer, environmental layer, what you end up with is a highly contextualized time line of what someone has been doing throughout the day and in which context. This is really, really important because once you have this, not only can you infer what they are currently doing, you can also start predicting what they want to be doing next. If you have the ability to infer current context and next context, then you can put that kind of intelligence in every single device you interact with so that they can decide what is the best way to interact with you: should they interrupt you because it’s urgent? Should they do it for you? Should they work together in the background so that they will do something else? Whether it’s a thermostat or a watch or a bed or anything like that, eventually all of these different devices will be able to figure out your intentions and anticipate everything you want to do with them. This is the real promise of context awareness.”
This is the vision of Snips and it’s certainly one that we find both exciting and necessary for a future where there will be many, many more connected devices than we have today.
Snips’ first product is a step towards this vision. They have built a mobile application that predicts what you will want to do next based on your current context.
Rand very humbly admits “This is just a first step towards this long term vision, and it’s probably not going to be the right product. I don’t know. To be frank, I don’t care because what’s important is that we actually try. If no one does, then we’re going to end up being enslaved by technology. We’re going to end up spending our days interacting with it instead of doing things we actually care about.”