 Hey, everybody, I'm Mike, I'm based in Berlin, where I make experimental games and interactive art installations, a lot of which use location and GPS and where you are in the world in interesting ways. So one of these, this piece called Computational Flinner that came out of my research at the MIT Media Lab, it's a generative site-specific poetry walk. So as you wander through Fort Mason in San Francisco, a poetry robot reads you neural network-generated poetry based on where you are. So you walk by the cannons and you hear pseudo-nonsense poems about war. You walk by the waterfront and you hear poems of the sea, that sort of thing. You can get it, it's free in the iOS app store, but that's not what I'm here to talk about. I'm here to talk about this sort of interesting problem that I found, which is, so if you've ever done anything with GPS, you might have run into this problem known as urban canyons, where GPS in cities is really inaccurate. So on paper, GPS is good to like 15 or 20 meters of accuracy in the average case. In cities, that can often go up to 100. It's really, really not good. And we'll talk about urban canyons in a few minutes, but the relevant thing here is, I did not have that problem at all. I was getting two or three meters of accuracy, which is totally unheard of, and it made my life a lot easier when trying to write the code to make this thing work. Whoa! There we go. Yeah, but it made me wonder why and what the heck was going on. And so it's something down this large rabbit hole of, how does your phone actually know where you are in the world? And spoiler alert, the correct question to be asking is not how does your phone know where you are. It's not even how does a computer know where you are. It's how have humans throughout history known where we are in the world. So to pick an arbitrary point to jump to, which is not the earliest time that people have known how to locate themselves, let's go to the age of exploration. It's the 18th century, you're a mediocre European white dude trying to conquer the world. And you really need to know where you are at sea. So latitude, like how far north or south you are, that's really easy. If it's noon so the sun is where it's highest, you measure that angle, you do some pretty basic math, consult a lookup table, you know how far north or south you are. Longitude, how far east or west you are is a way harder problem. Again, totally solved problem on land. People have known about this for thousands of years, but everything that worked on land didn't work at sea. The best we had was what we called dead reckoning, which is really straightforward. You say, all right, I know where I am. I know that I'm going at this speed in this direction. I can draw a line on a map and I should be here. That doesn't work great. It is so bad. So during the war for Spanish secession in 1707, there was a single event where the British Navy lost about 2,000 soldiers. This was not a battle. This was four different ships that were communicating with each other and each had their own navigators trying to figure out where they were. And they all said, yeah, we're totally nowhere near these rocks. And then all four ships crashed into the rocks. So every single Western European nation that was doing this whole seafaring thing put out huge cash prices to say, whoever can solve the longitude problem, you'll throw money at you. And there are a lot of really interesting stories there and how it took about 100 years for this poor guy to actually get his money. But that's the story for another talk. The solution was this thing called the marine chronometer, which sounds really fancy and looks really fancy. It's literally just a clock that works at sea. Again, it turns out that's really hard to make. But what's more interesting is why does that matter? So if you know where it is, if you know what time it is where you are, which again is pretty easy, you know what time it's noon. If you think about how time zones work, all that really is is the Earth rotates 360 degrees every 24 hours. That works out to about 15 degrees every hour. So if you know what time it is where you are, which is pretty easy. And if you know what time it is in another known location, say by having a clock that is the same time as Greenwich, England, you can do some pretty basic math by subtracting the times and figuring out, oh, that's how far east or west of Greenwich I am. So again, this was really hard because this actual physical clock was a problem, but this is a known solution for a long time. I think Carodidus first proposed something like this. So we're talking thousands of years, but this turns out to be super important because this idea that yes, if you want to figure out where you are, distance from a known location, that's fine, but that you can use time as a proxy for distance. That is super cool and super interesting. So let's jump forward to the 20th century. It's the mid-1950s, middle of the Cold War, Sputnik has gone up. Some MIT scientists noticed something really cool, which is the closer Sputnik was, the more radio signals they got from it. There was sort of a doppler effect thing going on. And so that isn't directly how GPS works. It's not using doppler effect or relativity or anything like that. But that meant 10 years later when people are starting to figure out, how can we use satellites or anything to use these nice new computers and help them have us figure out where we are, that led to this insight that the same way you could use time at sea to figure out where you are relative to a known point, we can use that with radio signals. What if we had a bunch of satellites up in space? So GPS, the American system has 24 satellites, but GPS is the American system, other countries have their own. So you have a satellite that's doing nothing but broadcasting out time stamps and you have a device in your pocket that has very carefully synchronized the clocks at the same time. We know how fast that signal should travel, so if you can compare these time stamps, we can know pretty accurately how far away you are from that satellite. So if you do that for a satellite, if you assume the universe is two dimensions for now, you know you're somewhere on that circle that you can draw around that first satellite. You do that again with a second satellite. You now know you're in one of two places where those two circles intersect. And a third time, let's uniquely identify, yes, I am exactly here, give or take, whatever your accuracy is. And that's really all that GPS does. Like if you're dealing with three dimensions instead of four, you need a fourth point, or instead of three, there's some fuzziness, you can actually use the Earth as one of those spheres if you want, if you don't care about altitude. You tend to use four anyway, because there's a lot of complexity that I'm alighting over in terms of how do you actually keep these clocks in sync. But that's really all GPS is. There's not a lot to it. But so if we jump back to urban canyons and what was going on here, it starts to make sense that if GPS is based on this idea of you have a line of sight radio signal that's coming in from space and you're going to measure exactly how long that took to send, if you're in an area with a lot of skyscrapers, you don't have that line of sight. That signal is not going to come straight to your pocket. It's going to bounce off of skyscrapers. It's going to make it delayed when all four signals are delayed, the algorithm in your pocket is not going to know what to do. So that explains why computational funer was probably better than it could be. That doesn't explain why it was so good. And I don't have an exact answer for that. But I guess we can see here computational funer is in Fort Mason in San Francisco, which is that park there. I'm realizing using a laser pointer doesn't work when there are three screens. But yes, it's in San Francisco, but it's in a not particularly skyscraper heavy area. It is also right on the waterfront. So you have this beautiful line of sight to the satellites. So that solves that problem. But what's also probably going on and why it is so darn accurate is today's GPS is not the original version of GPS. There are a lot of really fancy things that are happening, whether it's using Wi-Fi and cell phone towers to give you more accurate data, whether it's using essentially caching servers to get down satellite information, whether it's having ground stations that can send hard-coded in correction data. If there's any place that's going to have a lot of these bells and whistles, San Francisco is probably going to be well equipped for it. And so I could dive into why all that is and how that stuff works, but I want to jump over one last time with my last minute or two and talk about even farther forward than GPS, which is indoor location. Because it's really cool to me about all of this is all this stuff I've described is very high level. How can we take your distance from a known location? How can we use time as a proxy for distance? How can we use radio signals as a proxy for time? And so that holds true for things that aren't satellites as well. So if you've ever touched Bluetooth low-energy beacons or Wi-Fi-based indoor positioning systems, they work exactly the same way. Instead of a satellite up in space, you have a little Bluetooth thing on the wall that your phone is reading the signal strength. It can use that to determine distance. If you have a whole bunch of them, you can try to iterate the same way you can try to iterate with satellites. And they have the exact same urban canyon problem, where instead of bouncing off of a skyscraper, if a human being walks in between you and the beacon, that's going to attenuate the signal strength to the point that your phone doesn't know what to do. And this works the other way around too, which is super cool, that more modern indoor location systems use fingerprinting and other ways of generating data to use a probabilistic model, instead of just raw signal strengths, to more intelligently figure out where you are. And so if you're asking, can you do that with GPS, like, yeah, it turns out you can. We were just published this white paper a month or two ago about how they're doing just that in order to more accurately figure out what side of the street you're on. They're doing some really out there stuff with 3D modeling, where buildings are in figuring out what satellites you can't see to try to figure out where you are. So to me, that is what's really cool, that I set out to solve this problem of, how does the GPS chip on your phone know where you are? But it turns out all of these fundamental technologies, it's relatively simple math that just keeps carrying forward and that the exact same underlying concepts that drive satellites in space can also drive beacons in your art museum can also drive sailors trying to figure out where they are a couple hundred years ago. And that's all I got. Thanks.