Partnerships Lightning Talk, Jubal Harpster – Transcription

Okay. Can I ask people to get seated for our lightning talk session of the afternoon? And the final session of the day too. Thanks for sticking around. Okay. We have seven sessions today. So we’re going to do the same thing as we did before if you haven’t been to any of the lightning talk sessions. We’re going to do them backtoback. Seven times, five minutes. 35, I guess. And then have time for a group Q&A together after that. Please stick around to the end. Also the speakers. I’m going to be in the front and raising my finger if there’s one minute left. That’s all First up is Jubal with here comes Microsoft again, I think. So let’s get you started. Hello? Okay. Thank you. I’m Jubal. I work at Microsoft, I’m on the big maps team. I’m relatively new for Microsoft. About a year and a half. When I tell people I work on the big maps team. Two responses, the first is does Steve Coast work there. And the second is I thought they sold you to Uber. The team is alive and well. It’s Microsoft again. We have been here before. So I’m really going to give you an overview around the thinking of OpenStreetMap in the company and challenges about introducing OpenStreetMap in a big way. And then sort of what keeps me up at night. So just to give you a little bit of context, we have a long history of mapping at Microsoft. And if you’re old, like me, you have been through a lot of this and seen the developments. It’s been stable for a few years. Lots of good customers. And things are just sort of humming along at a reasonable pace. And to give you a better sense of actually where I sit within the organization, I work right the at front of the location data feeding into the data store. So some was things you see on bigmaps.com, routing, geocoding, reverse geocoding, rendering are the byproducts that come out of the team. The internal customers, Skype, Windows, Excel. A lot of the requirements from our team come from deep within Microsoft. And each team has a legal team, they have their own requirements and customers. We have to be cognizant of the data in the top of the pipeline about the tens of thousands of customers we could potentially impact. Some of the contributions over the year. I’m sure everyone is familiar with the Bing imagery. And around 45, 46% of the edits have source imagery. This is great. And DigitalGlobe to helping to create the ecosystem. That’s awesome. We just contributed almost 10 billion footprints this last year with the high attribute. Those are license the under ODBL, ready for import for the OSM community. Certainly the import process has to be followed. But the data is out there to use. We have this great internal program called the Benevity programming. Anyone spends an hour mapping outside of work, Microsoft contributes 25 bucks. We run the missing maps in the Bing org and spread it across. We’re continuing to scale this out. When I first came across open data when I showed up, it was about a year and a half ago. This is the general reaction. This is fantastic, we love it. This is going to be easy. Let’s just go for it. And then we run into this. So this is what I put in front of the Microsoft legal team about how we’re going to navigate this set of Wikis, these pages, all these talk forms. And they got  they usually end up like this. But for OSM, we didn’t just say the ODBL license is too restrictive. But we’re figuring out what we can and can’t do. One of the things is comply with local laws. That’s a problem. Some places have rules how you can display data and manage it. If you’re in a China, Taiwan is not a country. So we may have things that don’t impact the armchair mappers, but serious for Microsoft. No bulk updates. All the street names have to change sometimes. We can’t just do that in OSM. We can’t conflate data. We can mix and match data and engineer around the problem, but can’t actually combine the data to create new and different and excites data sources. And we can’t sue someone. Nobody to sue. I can’t sue Kate. There’s no money to be gotten there. If somebody sues us, we can’t turn it around and sue somebody. That’s a problem. And can’t pay money to make it go away. Can’t license the OSM database for $20 million. There’s nobody to pay it to. The data is a resource for everybody. So some of the things we’re sort of working on. We are doing prototypes around deep learning models using our internal GPU cluster. CNTK, a built-in framework for deep learning. And doing extraction. And can we create a building footprint which is indistinguishable from a human edit? We are doing prototyping with side by side. So one of these sides is Bing maps, one is OSM in the Bing style. Let you figure out what it is. So the things that keep me awake, vandalism. We are terrified of vandalism. We rap a simple filter and came up with data in the database for years and nobody noticed. And we’re worried about sort of the overt vandalism, but subtle forms. Changing default language tags and moving boundaries around. Can cause a real problem for Microsoft. Infrastructure. I worry about infrastructure all the time. So for somebody like Microsoft, we have to worry about directed, organized, targeted attacks. And would the OSM infrastructure be able to handle it? We’re about scale. See users going up and see the percentage of users actually contributing to around 1%. But what happens if this goes up ten times? That’s an issue. And finally, we are worried about GRASS. We don’t know what it is. It’s not a park. I can’t map it, Central Park I can map, but the grass is blank. Those are some of the things. That’s my five minutes, I’ll take questions at the end if there is any. Thank you very much. And there’s seven of us here the last couple days. Definitely try to find us if you have some more questions. [ Applause ] Thanks. I think that was the very definition of lightning. Okay.