The future of search and computing, or why and how I joined Snips.ai
An introspective view on a tech challenge worth tackling
Prologue
Not long ago, in this very galaxy, forces at play brought me to the end of an entrepreneurial cycle. Yet, contrary to popular beliefs, exits for a founder are neither frequent, easily planned nor smooth. In this instance, I was so deeply involved in managing the M&A process and organizing my personal exit, that I found myself without my next move.
It’s not in my nature to be sitting idle, so I decided to become systematic about looking for my next occupation, the one that I would allow to capture the best of my professional attention for the decade to come. As part of this non neutral search, I embarked on a social experiment to partially crowdsource my future. The experiment provided a tremendous boost in my search, allowing me to meet up to 4 people a day for the best of one month. So a big thank you to all who participated!
Casting aside the almost-by-default creation of yet another venture for the time being, I entertained a wide array of scenarios, including joining established yet struggling corporations looking for their digital revolution, or pure players and other GAFAs.
My social experiment did more than help serendipity; it literally served me my future on a plate. Four distinct people summoned me to meet this fellow Rand. Sure, I had heard of him and his project — but had too little intel to form a precise idea about the venture. My contacts knew something I didn’t but knew me sufficiently to make the connection...
Flashback
Over 20 years ago, I dabbled in artificial intelligence for my final year thesis as an undergraduate in Computer Science. Back then, artificial neural networks were pretty underwhelming. Still, I was able to code and train a couple of encouraging POCs including a 4-word voice recognition engine, and a digit recognition one from a 5x7 grid — all coded using a mix of C, C++ and Objective-C on a NeXT computer. While fascinated by the perspectives of self-learning, enthused by the foresight of big thinkers like Marvin Minsky, I let the siren calls of a paid job lure me into financial software, and sadly ended my fling with AI right there and then.
Furiously talented team and a moonshot
And so I met Rand, Snips founder and CEO. 30 years old, and quite ahead of the curve. Then his co-founders Maël and Mika; highly complementary and equally talented. A few meetings later, I met early Snipsters Olivier, Joseph and Colas. A furiously talented bunch, with chops to eat data-science for breakfast.
Snips was already a couple of years old, acting as a Machine Learning research lab for large companies, and known publicly for this past work.
What my contacts knew and I didn’t: a moonshot was brewing at Snips. After a couple of years doing applied research, they had started focusing on a major problem to solve: making technology disappear.
To execute on the vision, Rand needed to raise funding and bring more “talent” onboard. To my surprise, the founders were remarkably mature and already quite advanced; a number of termsheets in hand, and ready to onboard people like me, and I don’t just mean older!
And so, I went through a grueling interview process (but don’t be intimidated if you want to join!), like the rest of the Snipsters now. The process went both ways. I had more than a few questions of my own. And it took more than a few weeks to sink in and project myself into this ride. Each team member also had to vet me as I wasn’t a neutral recruit.
On May 1st 2015, I joined as COO.
A few weeks later, we closed a massive round of seed funding. $6.3m with two lead US investors: The Hive and ENIAC Ventures, a few angel investors (Brent Hoberman, Dave McClure and Xavier Niel), and funding from BPI France, a French government sponsored investment bank. This pre-product seed round is a testimony of two things: that US funds are looking at EU startups and talent as potential global players, and that France may provide just the right educational background to tackle hard data-science problem. The European Startup Revolution?
Most importantly, I’m a strong proponent of what we can build today out of France. As a non-professional lobbyist and co-founding member of France Digitale, I believe the conditions are ripe to take advantage of a relatively untapped talent pool as long as the ambitions are global. I’ve written a piece in French on this topic, taking cues from the tiny Startup Nation.
What does it do?
All too often, technophiles and early adopters want things to be overnight or OTA delivered the moment they hear about it. A lot of impatience all around (as if all overnight successes actually were overnight!). The consumer imagination has been empowered by the smartphone; by carrying a supercomputer in their pocket, supplemented with over 1.5 million distinct solutions to life’s more or less futile problems. As a matter of fact, many consumers and their dogs have also become would-be-app-makers, with finely chiseled ideas about what this or that app might do for them. Unfortunately, as I’ve highlighted before, the app stores are not long tail, quite the contrary. Unbundling can only go so far! The charts have been dominated by a few winners, leaving them to capture most of our idle and mobile time: Facebook with at least 25% of that time (or your battery), and WeChat, the one app to rule them all in China. With all the sherlocking going around, mobile OSes try to remain dominant with a one size fits all approach. The platform war is alive and well, and yet, network effects ensure a near monopoly position for each of the dominant players. Once they dominate, there is a reverse incentive to reduce their footprint; often coercing end-users into their sprawling offering.
One thing for sure, Snips is not your next time swallowing social/media platform. In fact, Snips very much wants to be a different kind of enabler and get out of the way… #redux #counterpoint #platform
The problem we’re trying to solve
The main problem is rooted in technology itself. While computer science and computing at large have undoubtedly delivered improved productivity, contributing to making the world and mankind more connected than ever and participating in cross-disciplinary advancement (medical research and genome sequencing for example), technology has turned generations of consumers into guinea pigs for what is to come, and garbage producers with planned obsolescence. Yes, software is eating the world, but also eating at us, by capturing our fragile attention and stealing precious time, a scarce resource. In his TEDx, Rand makes a point with the phantom vibrations that we sometimes feel; Pavlovian’s false expectation for our next sugar rush!
What’s worse is that end-user attention is mostly redirected towards the futile, easy to consume low hanging fruits, instead of promoting education, equal opportunity for women globally, reduction of our dependency on fossil fuel…
So there. Consumer technology gone astray and turning mankind into click-bait is a big problem to solve.
Bridging the gap
Too big, too vague? Let’s break it down.
To make technology disappear, we need a change agent, one that will accelerate the transformative process of the UX and focus point on a daily basis.
In fact, it doesn’t take Spike Jonze to imagine what happens when you put connected devices everywhere on and around people: the interface fuses with the user. The User becomes the Interface. The interface disappears, and becomes ambient, pervasive and ubiquitous. Ubiquitous Computing is not a novel idea, but it’s surely within reach now. #ubiquitousComputing
To do this, devices need to be somewhat connected to each other and to the cloud (#IoT), but most importantly, somewhat aware. To become aware, devices need what we call Contextual Awareness, a branch of Artificial Intelligence that combines pretty advanced technologies such as NLP, Machine Learning, Signal Processing and all sorts of cutting edge data science (including new generation of neural nets). AI can take the complexity around us and reduce the scope of interaction, and search (more on that below).
This is precisely what we are working on. A layer of technology to embed on any connected device. And, yes, we’re working on bringing this to so-called smart-phones, as a first step, mostly because they are currently the most ubiquitous connected devices.
What about Privacy?
Privacy is another problem, or at least one that has become a major issue; political, ethical, commercial, and fundamental. Currently, three forces at play are pulling in their own respective directions. A real conundrum.
To be fair, we’re not setup as a charity; but we don’t want to put ourselves into a position where our end-users become the product; a sure way to slip into advertising and data reselling!
Hence, our approach is radical. We call it #PrivacyByDesign. We posit that by leaving your personal data on the device itself, instead of on the cloud, we break free from the conundrum above. If we don’t have the data centralized, we can’t possibly abuse it.
For more advanced computations that need a sort of server-side handshake, we are resorting to a number of known best practices like record anonymization and decay, as well as new techniques like homomorphic encryption. Work in progress, but exciting and bleeding edge stuff!
So, what does it do?!?
We’re building an iceberg structure; so the visible part can only leverage a fraction of the underlying tech in the short term.
Nevertheless, we are determined to ship early and break things. In fact, a few thousand beta testers have layed their hands on preliminary releases on both iOS and Android.
For the time being, Snips is an AI for smartphones that analyzes all of your data to suggest relevant apps and content within them. We are currently focused on mastering location context awareness. What this means is that the data-science model knows where you are, where you’re headed next, and not just the GPS coordinates: it knows you’re heading home or to the gym as part of your daily routine, or it knows where your next rendez-vous meeting is relative to where you are and that you probably won’t make it given the current traffic conditions and your typical transportation mode.
Earlier this summer, we released a keyboard extension on iOS built on top of that above layer. Bare-bone UI: a keyboard without keys mind you, just one button, the one with the address you need.
We recently released a full-fledged iOS app that takes that location awareness to the next level and deeplinks to your next logical app; so that you don’t have to manually launch 3 apps, copy-paste between them and finally give up to hail for a taxi.
In parallel, we released a radically different value proposition on Android. Not an app, a new gesture that integrates deeply into the OS. One that has all the features of the iOS app, but that can also analyze the content on your screen to provide relevant deeplinks for it. We’re particularly proud of our advancement here, especially compared to Google Now on Tap on Android M: Snips kicks ass on location based processing, is compatible all the way to Jelly Bean, has #PrivacyByDesign, and plays nicer with 3rd party apps.
So that’s what it does. Today. Present tense.
Tomorrow, we’ll be adding more data verticals (location first, time and people next), and more platforms (too early to share). In fact, we take pride in building specifically for each platform according to its own specifications and how people really use them.
Bear with us: it’s just the beginning. Our team is 30 people strong now, half data-scientists, the rest no less data-activists. We’re working on this everyday, and quite possibly in our sleep, though we don’t have the data to prove it!
The Future of Search
The future of search is quite simply: no search. By that I don’t mean removing the findings and the discovery. No I mean the act of searching as we know it from the desktop web.
Given context awareness and mobility, most of the findings and discoveries can and should be inferred from your personal knowledge graph, and presented as readily actionnable items, reduced to their simplest form, the ultimate transaction. Triggers and interfaces are quite irrelevant ultimately, but won’t go away short term.
Sample use-case and Business value
Here’s an illustration that isn’t too farfetched: stepping outside your office, and considering that it rains, that you are both a Uber and Yellow Cab user, a vehicle either black or yellow should pop to the curb and bring you home. The act of entering the vehicle is both interface and validation. Time saved. Efficient transaction to go from A to B.
In a nutshell, we’re building value at scale, looking to impact billions of lives, repeatedly saving fractions of time, and enabling seamless transactions otherwise missed opportunities. A technology layer powered by the kind of AI that is there to help and not annihilate us! Exciting tech with a humane goal.
Epilogue
You’ve made it until the end? Bravo! Long form reading is difficult with all those interruptions isn’t it?
TL;DR? Don’t take my word for it anyways. Sign up for the beta and/or consider joining us. Opportunities to work on technology this forward looking and this ambitious are rare, oh and did I mention the awesome team? I certainly made my choice.
And Happy New Year 2016!
