 Thanks a lot for coming. I'm Greg Conti, and I'm here to talk about satellite imagery analysis, and this springs from I've been doing work in analysis of visualization of network activity and security data, as well as web-based information disclosure, like what you disclose as your user computer to large online companies. But in a previous life, I did satellite imagery analysis professionally, and I went out and I tried to apply those, find some information on those techniques that I could apply in my own work, you know, analyzing images, and it just wasn't there. And at the same time, if you think about it, you have tremendous power, power that only really nation states had about 10 years ago in the form of things like Google Earth and an online imagery analysis tools. So I thought I'd go over some of the tools and techniques that should help you increase what you get out of the imagery that's available, because I've worked with some truly amazing people, like you'd see it capture the flag here, that level of expertise, but only with imagery. And it was amazing to see, and there's some lessons learned like hope I can pass on so we can use this power in appropriate and effective ways and find some of those unanticipated uses that are really cool and really serve the common good. So with that, let's get started. So first, I'm here as a private citizen and not a representative of the government. To understand imagery analysis, we have to understand, you know, the implications of it, because now that the world has access to these free tools, what does that mean? And this might be a bit of a review for some of you, this piece of it for those of you that are really, you know, reading the news every day. But I think there'll be some pieces of interest, and I can guarantee you throughout the talk there'll be other things I'm sure you haven't seen. So as you study imagery analysis and try and, you know, get better at it, it's useful to know the historical context, where it came from and where we are today, as well as the capabilities of sensors, because the sensors, you know, it doesn't have to be on a satellite, it could be on an aircraft, and it really ought to be aerial image analysis or, I'm sorry, overhead imagery analysis in many ways. And the sensors that you see on the ground today could very well be in an aircraft, you know, a few years from now. I mean, it's just sensors and platforms can be largely mixed and matched. Then we'll get into some specifics about some analytic techniques, and then watching the watcher is the idea of what you're disclosing as you do this. So implications. We have unprecedented access to satellite imagery, high resolution satellite imagery that was just, you know, couldn't be, it just wasn't capable, you weren't capable of doing technically or at any time in the past. So we have this access and people are finding all sorts of interesting things. For example, apparently there's an abandoned pool on the roof of the Riviera Hotel. So things like that. I wish I could have announced that Friday night, because it would have made a good place for a party, I think. And of course, advertisers have jumped into the fray, making gigantic ads that can be seen from space. And then meta-advertisers jumped in, and there's actually companies that specialize. This company is called Artfield. It's out of Germany, and they make huge art so it can be captured in things like Google Earth. But it's also been used for positive things, such as disaster relief, hurricane Katrina. People used it to assess damage. And actually before and after, and I know at the time Google set up a special website, and they pumped very current information in there so people could, high resolution information so that they could take a look at their homes and their areas. And it was used for such things as insurance claims and the like. And people have been using the power of this in form of mashups by combining the data, the imagery with some data set that has a geographic component. So people like the Sierra Club here want to highlight the impact of a United States government land sale, so they made a mashup of it. People are using it to keep their governments a little bit honest. Speed cameras in Europe is what's depicted on this. And of course with this type of access to information, world governments and people in power haven't been particularly pleased. So one good example is that the Indian governments had suffered a terrorist attack and are very sensitive in their, I think it was their parliament building. And they wanted Google, they requested that Google take down some sensitive sites. So some people got caught with their pants down and some people got caught with their tops off. And this is today's technology now. You have to understand this genie is out of the bottle. Technology is racing along. Where are we going with this? Okay. And if you don't have access to the information, you know, just because it's being filtered at a major online company doesn't mean it doesn't exist. And then we're starting to see censorship, those people who've complained in various ways. And this is Sing Sing Prison, which is just north of New York City. It's where, you know, historically a famous prison where they send gangsters and things. And on the left you can see it looks like a big photoshopped out blob because it's been apparently sanitized. And then right across the street, very clear, crisp imagery. And the same thing at the White House. And this is, you know, the Washington D.C. kind of evolves over time and goes back and forth. But on the left you can see that the White House has been just replaced with brown tiles. And on the right you can see the White House. And it was interesting in the censored version that like the pool disappeared, as well as the roads and everything. So they just had this vague outline of the building. So this brings up an important point, though, that the censorship is obviously taking place. And to what degree do you trust this? I mean, it's probably true, but you have to always in the back of your mind question the assumptions. And then you're combined, you know, several people in the past, Amazon with their A9 service, and more recently with Google with their street view, have combined overhead imagery very seamlessly with street level views and drive vehicles around with cameras on the top. And they've captured people doing all sorts of mundane things, like buying hot dogs. But they've also captured people, I should have probably highlighted it, but on his left hip, you can see a big key ring. So he's probably not locked out, let's say. And they've captured people, I mean I'm sure there's prostitutes in there, I know they've captured people using drugs, anything and everything. Privacy implications of having this type of access is really amazing. And then you have to project in the future where this is going to go. And if you're going to do this, you're going to launch a new privacy-sensitive online service, make sure you capture an EFF attorney on the way to work, they like that. Particularly, there's a little back story here, apparently he smokes and didn't want his parents to know about it. And he was captured online or by the car, and apparently also by the A9 service, which is now defunct, so he was in there twice. And then, lo and behold, his picture disappeared from the service. So as we use these services, we have to consider that they may not be there when we need them. They may not be true, they may be sanitized, and they just, you know, not available. So where did this come from? That's where we are now. How did we get to this point? Well, aerial photography has been going on for a long time. The first aerial photograph was 1858, taken from a balloon in Paris, and it didn't survive to this day. Then you've seen balloons, pigeons, kites, compressed air rockets in the early 1900s. And the first photo taken from an airplane was by Wilbur Wright in 1909. And then World War I, World War II just accelerated that. So this is just a quick shot of the Paris photo, the oldest surviving aerial photo taken from a balloon. And this was a panoramic camera, taken from a panoramic camera hung from a string of kites of the San Francisco earthquake aftermath. Pretty amazing for 1906. And then again, World War II, they used to overhead imagery extensively. And here's where it starts getting interesting. Corona is the first photo reconnaissance satellite in the world, and it date back to the late 1950s, but went all the way through the 1970s. And it used film canisters, so they'd launch a satellite with film canisters, and they would drop them via parachute. And because it was so secretive, they didn't want these things landing on the ground, so they caught them mid-air by aircraft. Pretty cool. And those are the KH1 through KH6 satellites. And if you ever heard the term keyhole, that's where KH is derived. It was declassified in 1995, but it's not really common knowledge, and it's really interesting. And the NRO and the Air Force have some great information out there. The highest resolutions it had was 2 meter resolution, and 800,000 total images taken during that time. So the left is an artist's concept of the satellite, and they actually have pictures, real pictures of the cameras used, and the Thor-Agena rocket that was used to that class to push it into space. So just a sample from 1967. That's pretty amazing when you think about it from space. That opened up a whole level of access. I mean, in the past, a fence or national borders would provide a great deal of security from information collection like this. Okay, so we've taken a look at kind of where we've come from and where we've come from and where we are now. Let's take a little bit about a look at sensor types, because that's really a key component of this. And there's a wide variety of different sensors. You have them on the ground, but each has unique characteristics. Most of what we see in things like Google Maps and things like that is electro-optic or visible. Think your normal camera. But there are many other different types out there. Radar imagery does things like, notice on the left there's no clouds. So the energy passes through the clouds reflected back. And you can see that oftentimes it'll pass through vegetation, so you can see the lay of the land more so. And then other types, depending on the band that you're looking at, you can see thermal people. You can tell temperatures of things as well as infrared. And then multi-spectral just looks at multiple bands. People in the corona satellite had this capability. You can generate stereo pairs, stereo data. And then you combine them together and get a 3D image. Okay, this is a key piece of imagery analysis, or just a key component. And you hear these terms bandied about. The resolution of the satellite in terms, usually in terms of meters. It decreases over time. And you can see the image on the right is 10 meter. And it goes down to 1 meter imagery. What that means is if you took a 1 meter white square and laid it next to a 1 meter black square on the ground, it would be visible. But if it was less than that, it probably wouldn't be visible. So that's what it means. Okay, I have a quick demo. So whenever you have a sensor, you have to make design trade-offs as far as you're constrained perhaps by the bandwidth you have to communicate with the sensor or the amount of power that you have. And how many samples do you need? How crisp does the image need to be? Do you need more of them? Of a lower resolution or just a fewer higher resolution? So for example, if you wanted to... If you wanted to sample an image, it's called spatial sampling. So you're going to take some number of samples, each one of these squares is a sample, and you think of it as equating to a pixel. So this is actually a contest. If you can tell me this is going to evolve and get clearer. So the first person who can tell me what it is wins. So seven times ten, 70 samples, very coarse-grained, but the storage and processing requirements are very low. And then over time, it gets sharper. That's good. Stormtrooper it is, yes. I think the people in the back have an advantage because your brain actually can fill in better than the front. They just see the pixels in the back, they can see. It kind of, your brain fills in. If you have one in your room or something, it's probably even easier. I have a poster, so okay. So as you go, you can see getting progressively clearer, but the size is going up, the size is going up. So you're paying a price for the increased quality. In any type of sensor technology, you're going to have to make imagery sensing technology, you're going to make similar decisions. It's also important to consider that there's a point where you exceed the capabilities of the human perceptual system. So there's a point where you don't need any more, at least from the human eye perspective, because you're not going to see anything. That's how it's... Yeah, I mean, that's what we're doing. There's a certain point you hit where you don't need any more because your eye just can't resolve a difference between one and a higher resolution. So the other piece of this is those are the samples. Now how much data per sample do you capture with the quantization? So if each one of those samples was just one bit, again, very little overhead, but if you think of it two to one just gives you two possible colors, like black and white. Two to the four bits, you get 16, but you're going up in size significantly. Eight bits, 256. And 24 bits, which sometimes you hear as true color, which by most estimates exceeds the human perceptual system with a little bit of margin of error in there, and you get much beyond that and you can't tell the difference. So there's a reason why 24 bits is generally the limit. Okay, that's it for that. Okay, so we've just talked about sensors and some of the design decisions, and it doesn't have to be on a satellite or a plane. It can be handheld, but it does relate to the whole larger problem. U2s have been used, aircraft have been used extensively, very famously actually, and this is U2 on the left in a missile launch site in Cuba. I believe during the Cuban Missile Crisis. And today we're seeing a whole other set of platforms being used on manned aerial vehicles. This is the Global Hawk. And what's interesting is that there's a great deal of open source publicly available information on these different sensors. This picture is an example. So you can take a look at the, they make a lot of this readily available. So if you're interested, if this is your area, you can go out and you can download all sorts of stuff and find out, well, where are the capabilities of this aircraft? How long can it fly? What altitudes? And what type of sensors does it have? Okay, let's talk about space because there's some special characteristics that apply for satellites that don't apply to air-breathing platforms, things like aircraft or you on the ground. When she launched this thing, it's up there, perhaps for a long time. So do we have any amateur radio satellite type people here? A couple, okay. So the amateur radio community has actually been all over satellites for a long time and there's a great deal, there's a great wealth of information out there. Satellites are obviously objects in space but they're in free fall. So they're falling around the earth and it's a gravitational pull of the earth that's keeping them in orbit around the earth. And also you are in like the Leo orbit, which is low earth orbit, very quick. So you're talking like 90 minutes. And as you get farther away, the orbits are slower. And if you get out to 22,300 miles, the orbit is very close to the rotation of the earth. So when you point your dish up to one point in the sky, it's because it's at that orbit. Another way to look at it is there's geo, geosynchronous, geostationary. Generally geosynchronous is the one that I think people like a little better. Each of the orbits implies a different use because certain things like putting a camera on a camera 22,300 miles away is a lot to ask. So different things will be used for different missions. And this is from an international security magazine. And the elliptical orbit there is called the Molenaya orbit named after the satellite that first used it. What's interesting is that it moves very quickly where it's close to the earth, it whips by, but it has a long dwell time, the long oval, so it goes up into space and has a long dwell time. So you have a great deal of visibility a little bit closer than you would get over a geosynchronous orbit. So Keplerian elements are part of a mathematical model and you run into these, sometimes they're called KEPs, but when you have them, it mathematically defines the orbit of the satellite. So you combine that with where you are and the time and you can tell where it is. So this is actually, I have a link on the slide if you're interested, it's pretty interesting, kind of technical, but it allows you to know where the satellite is and so where to point your antenna. And there are tools out there then that have this built in. This is one that I've used and I like and it has actually a reasonably powered demo version and it has the KEPs built in for a large number of satellites. So you can just select the satellite you're interested in and find out where it is. Pretty cool. Okay, so transitioning from kind of the sensors and the constraints that you're dealing with, actually one final thing about satellites is if you heard the term like from Star Trek, let's put the ship in a parking orbit. You can't just fly satellites around lightly. I mean they have a limited amount of fuel. So you can't just put one somewhere and turn on the thrusters and stay where you want. I mean, you're talking minor station keeping constrained by the laws of physics. Okay, so these are some basic steps in analyzing imagery. In years I'll get into a second. But what you're trying to do basically is orient yourself to this new image. Locate things, locate it on the Earth. Find things of interest and identify them by class. Like oh, that's such and such. And you might be able to identify them uniquely and I'll show you. Like that's a specific, you know, of 757s. That's a specific one. And along the way you want to build domain knowledge because if you're analyzing certain places, you know, like for me I could take a look at a military base and tell you every little component because I've lived on them for a long time. And I'm sure if you work in a port, you know, you can tell what all the things are. So you want to build that type of domain knowledge of what you're looking at if you're not an expert to begin with. So the idea of the nears is it's a rating from one to bad, nine to best that allows you to quickly classify an image. And there's the link, I have a link here that tells you what's reasonable to expect to be able to pull from that image. And sometimes it might be surprising, but with a little practice, and I've, like I said earlier, I've seen some amazing people that do this, you know, they've done this for 20 years. They can really tell a lot. So you're constantly to push yourself. So nears one, the basic idea, you know, even if it's coarse-grained imagery, you can still see things like ports and that's the nears one is the red square. And then it gets a little better and you can detect hangers at airfields. And then getting in, you can start seeing houses at nears three. Nears four, you can tell aircraft by type. And there's actually a table, like each of these, there's like 10 different types of things you can see at each level. But I tried to choose ones that were generally, you know, reasonably applicable. And then it gets closer. Just for, that's not the best photo. So the red square is a house, or a large building, a farmhouse. So going a little bit more, you know, identify a ladder on RV is nears seven. So what we're seeing today is probably nears six maximum resolution, maybe a little above, except, you know, street view doesn't count. I showed this to a friend ahead of time and he asked if number, if nears eight was taken from a satellite. He's been watching 24, I think. To the best of my knowledge, that was taken by a person standing there with a camera. And nears nine, in the red squares of shovel. So I mean, that's, like, you take pictures in the room, that's nears nine. So one of the first things you learn when you study imagery analysis is that you need to look at the images from the vantage point of the observer. Because you'll get really bad headaches if you don't. And, like, this is tilted, but you oftentimes see this, the images will be tilted in a certain way. Your brain has a hard time processing it. So what you want to do is rotate it. And I know the white corners make it look a little odd, but it's much easier to deal with cognitively if it's aligned as if you were the actual viewer. So when you're looking at these images, there's certain things, you know, some general things that you could be looking for. You don't go through all of them. But, you know, you're looking at these characteristics, trying to get a feel, what is this thing I'm looking at? I don't know. And context is kind of a vague term, but how is it used? What's in the surrounding area shaped? What time is it? What's the date? You know, is there power going into it? Is there any unique characteristics, that type of thing? So you want to orient yourself, and even at low resolution, you can see things like major highways, airport, which is in the rectangular square. And it's also useful to conduct comparisons over time. And I know that's kind of hard when you're using static imagery, using, like I say, Google Maps or something like that, but there are different services out there, and sometimes they'll have different data sets, so you might be able to do some comparisons. So this is an unclassified image that was released. And what they're doing is before and after of a strike. So in the red square there's a building, and in the red square there's not a building. But that's the idea. They use these things for battle damage assessment, and you can see these things over time if you're observing a site. And no one says you have to just solely use satellite imagery. You might have other types of information if you gather. This is interesting. Cultural indicators. So you're looking for things that are anomalous. Anyone from Germany? Okay. How popular is baseball in Germany? Yeah, it's not very popular. So you have an air base in Germany, for example, that has that. Okay. This is actually Ramstein Air Force Base, and if you see, and I'm probably the opposite's true, you see a soccer field or a football field in other parts of the world, things start looking strange. So you're looking for cultural indicators. Here's another cultural indicator. And if you know this, don't say it. The gray area on the left is a school. Okay. And the reason why I know it's a school is because some of the students pulled a prank. They were angry at their teacher, and they put weed killer in the lawn to make a giant phallus. So actually they had fixed the problem. They replanted. Well, apparently in that time, an overhead mission came by and captured it for the eternity of the large, yeah, pretty good hack, I think, for fifth graders. And obviously I can't cover all the different facets here, but it's just trying to push your thinking, get you thinking a little bit differently. But even the modest parking lot can tell you a lot. You can tell you, is it a weekend? If you're looking at something or is it a game day? Is it not game day? The number of employees can be determined perhaps by the size of the parking lot. The types of cars in the parking lot, depending on the resolution that you've got. One problem with imagery analysis is when you're looking at a single image, it's called trans-loading. The idea is, you don't know, are these people loading or unloading? When you have the still image in time, it's very difficult. If you have multiple images, it's easier. So that you don't know, was that guy climbing up the fence or down? You don't know those things from one single image. I thought, this is a really hard talk. There's a fine line of what I can do and not get drug away in shackles. And I really wanted to do like an airport. Well, let me just deconstruct an airport and go out and show all the different pieces and show where the security gates are and all of that stuff. And I'm seeing heads shaking. And I didn't think that would be a good idea. So I decided to do it, just discuss it using SimCity as a metaphor. Who's played SimCity? Okay. So what you're doing is you're building a place, part of the world, layer by layer, component by component. So you can do the same thing when you're analyzing an area. You just do the reverse. Actually, I went through the menus of SimCity to keep it unclassified. And all those things that you're adding, you can identify. So you can find the roads. You can find the bridges. You have to make tax the people and make money and build these things. It's the same thing. You just look for all of them. If you do this, if you go through all the listing, you really have an awesome feel for what's going on. And then if you go back to that other slide about anomalies and outliers and things, if something doesn't seem right, after you've kind of reverse engineered the city, and it doesn't have to be a city, you can do the same thing with other locations as well. This is a key concept, the idea of building a smart book. And when I did this professionally, I had, they gave us posters of like every known type of, let's just say pictures of tanks. And then so you could just refer to your poster and say, oh, that looks like a T-72 because it has N-road wheels and the things on certain parts of the tank or whatever. And you can do the same. I mean, whatever you are studying, it's almost certainly you can gather, there are images out there of it. There are certain characteristics of it, of aircraft, if you want to look at different aircraft. You build a little library of those and that'd be great for sharing too. You could make a PDF of these things. Whatever it is you're trying to identify, you collect images of different classes of it and then from that different things will emerge. Well, like say for an aircraft, the wing shape, the size, the location of the engine, the shape of the tail, those all are unique characteristics so you can start studying these things. So for example, I mean even antennas, different shapes tell you different things. So actually if you have domain knowledge, for the amateur radio type people, you build antennas and you can tell probably wavelengths of antennas from different images. Depending on the size obviously, the resolution of the imagery you've got and as well as the orientation. Bridges, this may be more of a military thing because we're worried about like driving tanks across bridges but in the number of lanes, transportation routes, what's the class of the bridge? How much weight can it carry? How much is the engineer to look at the picture or something like that? And this is just a little simple smart book for aircraft identification. It's out there and you could build really cool ones. Jane's is one of the world experts in this. Jane's defense out of England. And when I mentioned about unique characteristic, I found this photo on the web and I was intrigued by the N number. Well, you can, on the tail, and you can actually, there's a database and N number inquiries database and you can uniquely identify what aircraft it was. And apparently this was a United States Department of Commerce and it tells you that it was in Hangar 5 of the McDill Air Force Base and built in 1978. So kind of cool stuff out there, particularly when you can combine all the open source stuff that's out there with whatever area you're studying. And if you're studying helicopters, you might be able to go back to the manufacturer and they'll have all sorts of spec sheets. Like this is a MI-17 helicopter out of Russia, I believe. And then there's videos of MiG-35 at an air show. And I'm using military examples because that's kind of my comfort area. But whatever you're trying to do, you gather this type of information and start bringing it together. So part of this is about skill building and you want to raise your skill set to a higher level. One thing I've found is certainly in the military domain is the idea of museums. These things are out there. You can actually look at them on the ground. They have plaques telling you what they are. Usually a map when you're in a tourist center gives you a map. And also you can look at them from above. And it's also useful to try identifying them at a higher resolution. At a higher altitude, say you slide a little bit farther out and push yourself. Because you can't identify these things. Certain characteristics will emerge at levels that you really don't think are possible. In other places like graveyards of aircraft, there's some really cool stuff out there. Taking advantage of aircraft flights, if... You want to believe how hard it is to find an image of someone looking out of a window in an airplane these days. So I had to find an astronaut. You know, take a ride around a seaport. If you're interested in ports or things like that, take a ride around the seaport. Any seaport. Because they're typically the same. And this is a really neat site called Wikimapia. And it's this collaborative analysis of people you draw a region and it has on the order four million hotspots. And it's been around since middle of last year. And if you zoom in, I mean, I've zoomed in at places and it's surprising what you'll see. I've looked at it. It's reasonably accurate. And you'll find things that are just... I don't care how good an imagery analyst you are, but there was like a brothel listed on a site in an air base, I think, in India. And they're like, so-and-so is brothel. I mean, that's just not something... You have people probably on the ground giving you information about the location as well. It's worth checking out. I know one person actually, like, thought it was cool and they put their parking place. They marked their parking place on the map. And there's other tools out there. I found Google Earth visual guides at digitalgeography.co.uk that I thought were pretty cool. If you're, you know, trying to use the tool or teach someone else to use the tool, maybe students or something. I couldn't resist. I'm sorry. But the people in this room are some of the most security-conscious people in the world. And if you saw the talk I spoke about last year, you do have to consider that you are leaving footprints as you hop around the world. And what are those telling about you? You know, locations of your family members, employers, every time you use directions. You know, imagine your corporate headquarters with every set of directions that ever were emanated from it or from hotels and coming into it over time. I mean, it's really a sensitive issue what you're disclosing. So just as one example, you know, this is Las Vegas, and say you're using it, you know, Google Maps kind of extensively, and you're looking at these regions in particular, perhaps typing in the street addresses, you know, because you're home shopping. You know, that would be kind of visible. And say you looked at the yellow one five times and the green one ten times, and then you clicked the, you know, I think the forward this link page or something, and you got the link, and then all of a sudden ten other people on the planet all converged on that one street location. You've probably just disclosed your social network, you know, particularly if it was in, you know, a matter of a couple of days. So it's just useful to think about these things. So I have a list of sources of more information. You know, Google site-seeing is an awesome, awesome place. I already mentioned Wikimapia. Cryptome. He's done things that most people fear to tread. You know, he'll take people's, you know, Dick Cheney's fishing cabin and analyze it and combine all sorts of open source information. Some great tutorials out there, that remote sensing tutorial from NASA. You'll hear the term remote sensing, and that's kind of another related domain. Imagery analysis really involves, like, looking at the higher resolution stuff, but there are lessons to be learned by studying remote sensing. And then the Federation of American Scientists are really the only people that have imagery analysis specific stuff out there, and I got some of the materials from there, and I strongly recommend them. So with that, are there any questions? Oh, sorry. Yes? Okay. The question was what we can say about modern techniques for camouflage. I know there are military labs that look at this very closely, and they'll do tests to see, like, you know, the military army's gone through a different camouflage pattern. That's gone through extensive testing, because the idea is with camouflage, they want your eye to divert off the center mass of the body, so it's harder to see or harder to see you at all. And there's a field called Battlefield Deception, an area of study, and the idea is they have different things they can place on the battlefield that look like one thing or another, and to the sensors as well. So it's called Battlefield Deception. I would look into that. Yes? That's a really good question. I mean, we ran into that with Katrina imagery. You know, there'd be, like, devastated areas right next to, you know, houses, you know, standing upright. So this interleaving of images is a problem. I've personally found it difficult to date the images directly. Does anyone have any information on that? Yeah. I used TerraServer, I believe, just a few weeks ago to play with this. My experience has been slow. So what he's saying is that the TerraServer has dates that the images were acquired on. You said you could buy dated images through Google Earth. Okay. Okay, well, we can take any other questions in the... Oh, I'm sorry, yes. What he was just doing was talking about the different sides of the road as you drive from mainland China to what was it, Hong Kong, and how they could deal with that, and just kind of studied it. But I mean, which side of the road you drive on is a great cultural example to know if you have an image and you're not quite sure, that would be a thing to look at. You can find it and always use it in the background. Ah, so at intersections, yes. Okay, well, so the closing idea here is that this is... We have tremendous power right now. We can talk some more afterwards. It's kind of hard in a room this large without microphones. We have tremendous power to do good and just try and find ways to use, you know, this... In unanticipated ways, push the envelope. I mean, the mashups are a great technique of combining external data with an interactive map. So with that, I'll take any other questions up front, and I thank you for your time.