 I'm an astronomer, as you've heard, and I wanted to start by reminding you what data used to look like in my field. It wasn't long ago that this was cutting edge. This is a sketch carried out over a couple of nights by a man probably wearing a top hat. And we still study these things in much the same way. Of course, we take pictures, but the measurements we want to make are necessarily simple. We can't go and prod these things, and so we ask questions like, what shape are these galaxies? The shape of a galaxy tells you about its history. It tells you about whether it's collided with other systems. It tells you about whether stars have formed, and it tells you about how those stars are moving. The problem is that our maps of the universe look like this today. We have surveys that cover a million galaxies, but we still want to ask those fundamental questions about shape, and that's a task that computers aren't as good at humans as doing. We still have a human advantage in pattern recognition, but we have more data that we can cope with and more data that we can look at. And so the situation is one in which the machines seem to have run amok. We've got very good at building machines to collect information, to collect data, to get knowledge about the universe, but not machines which are capable of transforming that into a proper understanding. And we realized a few years ago that we needed to do something radical, and the radical thought was just to ask for help. We can use a crowd of people because of the digital world that we live in, because of the way that we're all connected. It's no longer just me and a student that need to look through the galaxies. We could call on a worldwide crowd of up to a million people and ask them to assist with our research. And it turns out that doing this has three key advantages. One is just the scale. When we put our first project called Galaxy Zoo Online, within a couple of hours we were doing 70,000 classifications of galaxies an hour. We weren't paying people. We just said, come and help us with science scale to the size of data sets. Perhaps more excitingly, these projects provide a place for serendipity in modern science. You find the things you weren't expecting. This is a visualization of Planet Hunter's 1B. The first planet known with four suns in its sky, discovered by volunteers via a website that we put together. We didn't even know to look for such a world, and yet they're able to find it. The third key advantage is speed. The red streak here is a distant galaxy, 11 billion light from which it's taken 11 billion years to reach us. Discovered within the first 10 minutes of a call going out on the BBC just a couple of weeks ago for people to come and help us. So 10 minutes from asking for help to making a discovery. And of course these problems and this solution isn't restricted to astronomy. We've run projects looking at all sorts of things. We've classified, for example, historical images of cyclones to try and get at this question of whether global warming is affecting extreme weather events. We've worked with cancer charities to speed the classification of cell images, like the one you see here. Turns out that when you go into a drugs trial or when you run a cancer trial, trained pathologists sit and look at hundreds of thousands of images like these. And we can show that for 90% of the images, and critically, we know which 90% the general public could do as well as the experts. And we can of course think about going beyond science. This is a picture of New Orleans just after Katrina, a satellite image. And you can see on it which buildings are standing, which roads are flooded. And I believe that that's information that we could quickly transmit to people on the ground because we have a workforce of more than a million people ready to stand by. But that task of moving to live images, because we want to intervene in the moment, requires us to work on this relationship between man and machine. If we want to get the best results quickly, we need to use both capabilities. And so we need to build a system, and this is what we're working on, which could take any given image and decide whether a human should see it or whether a machine and which human and which machine should do what. If we could do that automatically, we can construct a network of autonomous agents, each doing what they're most interested in the case of the humans and what they're best at in the case of the machines. The technology we're using comes from a surprising place. We've stolen code, borrowed code from our colleagues who are building robotic cars. They have code that decides which autonomous agent to listen to. And we're putting humans into that mix, accounting for that very human factor of boredom and being able to optimize our searching through data. So that's where we are. We have a platform that works. We have a million people. We have about 30 projects on what we call the universe. But we're looking for new and interesting places to use this capability, places to stretch ourselves, and places to stretch our volunteers who just want to help. And so my question to the room, taken very seriously, is how can we use this capability, the citizen science or this crowd to solve the world's problems? Thank you very much.