Brian Lent joined HERE this summer as part of the acquisition of Medio, which specializes in predictive analytics. Medio helped companies analyse reams of data in order to refine their recommendations. Now Brian and his team are integrating Medio’s predictive technology into location services for HERE to deliver maps tailor made to the moment.
Brian, tell us about yourself and where you’re from.
I started writing and playing computer games from the age of 12, on the Franklin Ace 1000. I went on to study for a BA in Computer Science, and then a PhD in Computer Science at Stanford.
Alongside that, I’ve always had an entrepreneurial slant. I started my first company – called Micro Revelations – at school, making games for the Apple IIe and the IBM PC.
At Stanford, at that time, studying computer games wasn’t really an option. You could go down the graphics route; or you could study gaming AI. I went down the gaming AI route, but after postings at Silicon Graphics and at the IBM Research Center, that ended up as my research PhD in the emerging field of data mining.
I established the MIDAS group at Stanford during that time to bring together all the people working in the space (the acronym stands for ‘mining data at Stanford’). Sergei Brin and Larry Page were also members. They went on to form Google, of course, which you could say was a by-product of my PhD!
I left Stanford to form a company called Junglee, which was the world’s first shopping comparison website, and included my PhD work on product recommendations.
In 1998, Junglee was acquired by Amazon which, at that time, only sold books. Jeff Bezos had a vision for the company that it would allow people to find, discover and buy anything in the world and he could see how Junglee’s technology could help him do that. I joined Amazon with the company, and worked with Jeff to build a lot of the data-mining technologies that let Amazon do such a good job at finding things you’re probably interested in.
In 2004 I founded Medio, which started off in mobile search, and then we moved the technology to become a cloud service for companies to be able to build their own recommendation engines; a kind of ‘Amazon-in-a-box’.
And now I’m helping to build that technology – which we call ‘predictive analytics’ – into HERE.
So tell us – using really short words – what predictive analytics is, and what it might bring to HERE.
Predictive analytics is designed to improve decision-making.
Today most of what we have is ‘descriptive analytics.’ They describe the state of something in the world. You look at the speedometer on the car, and it combines two or three sensors to tell you how fast you’re going.
Then we have ‘diagnostic analytics’. This tells us why something has happened. Your car stops running smoothly. You look at the dashboard, and the oil-light is red, so you know what you need to do.
Predictive analytics is about what will most likely happen in the future, by bringing together everything we know about what has already happened.
What is the fastest route to your destination? We take everything we’ve learned about traffic to date, evaluate all the trends within that and come up with a suggestion.
We aim to build a ‘data borg’, by adding and accumulating more and more information about the world into one system. It will be an index of the world.
Map training school
But the trick is in how you train the system. We form a closed loop for optimisation. It looks continually at the suggestions it makes. Did it work? Did you get to your destination on time? If not, then it adjusts the algorithm and tries again. So it gets smarter every day.
It won’t be just route-finding, though. We want to make the map into your concierge. You check into a hotel. The map knows my interests, tastes and how far I’d be likely to be willing to travel, and comes up with intelligent, personalised recommendations of where to go and where to eat.
Lastly, where is this going? Say five years down the line?
In five years, you’ll be able to ask questions you just can’t ask today. ‘What’s the coolest, hippest place in Seattle right now, for people like me, considering today’s weather?’ or ‘What’s the most popular place to go for people who own an iPhone 6?’.
By leveraging smart data to create location intelligence, we can start to explore the future.