 So, let's start with a question. What do cats as recognized by deep learning, satellite imaging and Siri, autonomous robots and Netflix have to do with discovering new medicines? In the next 15 minutes, I hope to share a story with you that helps to pull all of these threads together. So let's start in our break room. This is where we meet together as a team every day to have lunch. And on the wall are pictures of kids that remind us of why we're working on the problem we're working on. Each of these kids, science has let down. Each of these kids has a disease that in 2018 either can't be diagnosed or has no effective treatment. This young woman with the red hair is Ricky, and we've gotten to know her really, really well. She comes to a lot of our company events. And Ricky has a disease called neurofibromatosis type 2 that causes benign tumors to form on the nerves all throughout her body. And they can do surgery to remove some of these things, but at the end of the day there's really no treatment. And Ricky gives us a tremendous amount of hope, but she also tells us every time she visits that we give her hope. And that hope is completely audacious. And this is why. Today it costs more than $2.6 billion of R&D investment to discover and develop and market a new drug. And from the start of a program, when a company first decides they want to try and develop a medicine for a disease, 99% of the time they will fail. But the good news is that it's 2018 and we're landing rockets back from space in tandem so this must be getting better, right? That's not the case. Despite incredible scientists and companies dedicated entirely to finding new medicines, we're at the end of a 40-year trend of increasing costs. It's now 10 times more expensive to discover and develop a new medicine than it was in the 1980s. Compare that with the world of technology where we've seen a million fold decrease over the same timeline in storage costs. If you look at the number of new drugs that are approved every year, that's also stagnant. There hasn't really been a huge shift over the last 40 years. But the smartphone in my pocket has more compute than a 1980s supercomputer. So how is it that these two worlds can coexist today in 2018? They can coexist because biology is so incredibly complicated. This is a simplified schematic of a tiny piece of metabolism, just what we know about it, that's happening inside nearly every one of your cells every moment. And yet in the face of this complexity, the only tool that we've had at our disposal for the past 40 years has been a reductionist one. It's been our own human minds generating hypotheses one at a time, one disease at a time, scientists spending an entire career, decades, working on a single disease or trying to understand a single biochemical pathway. And that's the best tool we had, and it's resulted in some amazing medicines. But the low hanging fruit has been picked, and what's left are medicines that might be too complex for our own human mind to understand or discover. And so with this reductionist approach, we're left discovering treatments one decade at a time. So at recursion and at other technology first companies in this industry, we decided to look to the world of technology to see what we could learn. And we looked to the world of map making. We looked at satellite imaging, and in this space there's been a massive increase in data over the past 40 years. But this data by itself, these billions of images of the earth, aren't actually that useful. It's only when you layer on top of those images, machine learning and other types of sophisticated computation, when all of a sudden you can make that data actionable. I don't know if any of you grew up at the same time I did, but if I wanted to go somewhere, I had to open up a book, look for the place I wanted to go, find the address, call the place. Now I just ask my phone to tell me where to go, and I even get recommendations about different restaurants or other sorts of places to go in the way. It's incredible what's happened. So at recursion, we ask the question, what if we could map biology the same way we mapped the earth? And these are pictures of human cells, and they actually look somewhat like satellite images. But hundreds of millions of pictures of human biology by themselves, again, aren't that useful. It's only when you layer on top of that sophisticated computational tools that you can actually understand how to get from point A to point B. And so we've taken these images and applied computation to them to try and find new medicines more quickly. So how does this actually work? Well, this is a video from inside our laboratory where we have dozens of robots that are doing experiments almost every moment of every day. Right now, we're doing about 100,000 experiments every week. And by early in 2019, we'll be doing a quarter million experiments every week in real human cells, and all of those experiments will be being done by robots. That will generate about five million images each week. That's about 50 terabytes of new data each week, about a quarter million different questions in biology. And in the last 12 months, we've generated about 1.5 petabytes of these cellular images. And to give you a sense of context, that's about 60% as much data as all of Netflix holds for all of its movies and TV shows. So what do these images look like? Well, they look like this. These are different types of human cells, different disease models, different drugs, or some combination of those things. And you can see the extraordinary colors and textures and shapes. And all of that is rich information that's telling us something about biology. And so when you take hundreds of millions or billions of these images, in order to analyze them, you have to look to some of the newest computational techniques. And in fact, when we started the company, we were inspired by some of the work by Google, and I mentioned earlier the deep learning cats, some of the work they did to leverage deep learning to try and recognize cats in YouTube videos. And we've ended up in a slightly different space of deep learning, but we're using similar nets to try and build a representation of biology computationally from these hundreds of millions or billions of images. And if we can take all these images that we build in our own lab, enough of them, and we can query them the right way, we can start to understand biology at a different scale. And we have a new formula. A formula where you can simultaneously study hundreds of diseases with the study of each one of those informing the study of every other. Where you can evaluate millions of hypotheses and hand the very best of those to your scientists in the lab to validate. And we believe that if we do this, it can lead to rapid discoveries. So let me make it a little bit more real for you and tell you where we started, where we've been, where we are now. So at the very beginning of the company, we could do what we call radical empiricism. And we could ask basic questions like this one. Here's an image of healthy human cells. And here's an image of cells from a disease model of hereditary hemorrhagic telangiectasia. Try and say that one five times fast. And I've picked this because you and I can actually see a difference. You can see the disease cells look sort of long and narrow compared to the healthy cells. And you can then use your robots to add thousands and thousands of potential drugs to the disease cells and take a picture again. And you look for cells like this. These are actually cells treated with the drug that we're now developing for this disease. And you can see that they look much more like the healthy cells. And this is fantastic because we didn't need a hypothesis. We let the cell tell us the answer. And because we're using sophisticated computational tools, we can actually do this for images of cells where no human can tell the difference between disease and healthy. And that's the power of computer vision. But over time, as data starts to accumulate, you're able to actually see more subtle patterns. So as you start to do hundreds and hundreds of experiments across millions of biological conditions, you actually start to see how different drugs affect different disease models. Every line in this image is a different drug. And many of these drugs are ones that people take today for diseases. These in some cases are on the market. Everything in blue is what we know. Everything in gray is an interaction we found with some disease model that the scientific literature has never described. There's a lot of gray up there. And so when you start to look really broadly at biology, it's amazing how much more you can really find. But I want to tell you about where we are today, because this is where things really, really get exciting. And I'm going to start with Netflix. So this is my Netflix profile from about two weeks ago. I don't watch very much Netflix, but from just the few videos that I've watched, it's able to make some predictions about what I like. So it says I like dogs. Turns out to be true. I have an Alaskan malamute that looks a lot like the one in the picture. It says I like desserts, okay, maybe. It says I like cars. This is all true, even when I haven't watched videos about many of these different things. And it's because they have this massive data set for more than 100 million users, which they can compare me with. And we're trying to get to the same point in biology. And so I'm excited to share with you today this animation, which is real data from a subset of our diseases and drugs that we study. And what I'm showing you here is a high-dimensional representation, latent space for the purists in the crowd. For the rest of you, you just need to know that this is an image that shows different diseases and different drugs. And if they're near each other, it means that they're really similar. And if they're far from each other, it means they're really different. So in red, we have different models of disease. In blue, we have a couple hundred different drugs. And when we start to actually look at this data, we see some interesting patterns. So in red is a specific disease, and in blue are the drugs that end up making that disease better. And you notice that they're highly enriched for the opposite side of this representation space. And that's super, super useful because what it means is that we don't have to do every experiment anymore. We've generated enough data already that we can actually predict before an experiment starts what drugs are likely to work. And as we keep learning and keep adding data, we'll get better and better and better at this. Maybe in 20 or 30 years, there won't even be a drug discovery or development process. Maybe we'll be able to make predictions for each person immediately. Ultimately, the goal is to be able to build a map of biology. And we're doing all kinds of cool stuff. I couldn't help. This doesn't really fit in anywhere else. I couldn't help but show it. This is some work we're doing to try and predict toxicities and side effects of molecules. And these are actually 300 small pieces of heart cells that we've grown in our laboratories that are all independently beating and we're watching that using calcium imaging. And we're trying to be able to do this across tens of thousands of these little tiny mini hearts all at once so that we can identify whether or not a drug that we're interested in might have some effect on somebody's heart that could be really bad. And these are the kinds of things that technology is enabling today that weren't possible five or 10 years ago. So if we can put all these pieces together, if we can build a big enough data set that's clean enough, we can ask the right questions. Our hope is that eventually we'll be able to make inferences and predictions about something really, really important to all of us, which is new medicines. We might be able to predict even that a chemical that's never existed in nature, it's never even existed, might be the perfect chemical to treat some disease. We might be able to predict things like biodistribution or solubility or all these other things that make it cost $2.6 billion to develop a new medicine. So that's something that we're incredibly excited about. But let's make it real. So I told you about a whole bunch of things that at least I find super cool. At the end of the day, none of it matters unless we're getting medicines to patients. So I'm happy to share with you that as a five-year-old startup, we now have our first drug in human clinical trials. So we're actually testing a drug in humans in the US under the supervision of the FDA. That's a huge milestone that a lot of biotech startups never even get to. I'm also happy to share that we are just finishing the in-licence of a second drug. And this drug will now be used actually to treat the same disease that Ricky has that I showed you earlier. We're super excited about this. Beyond that, we have two more drugs that are on their way to the clinic. Before that, we have about a dozen drugs that are in animal testing right now and more than 30 programs that are earlier stages. And over the next couple of years, our hope is that instead of taking one or two drugs to the market every few years, that we can actually take dozens and dozens of treatments to the market with our partners, like Sanofi and Takeda, who we've signed deals with. And we're currently working on more than 500 different diseases using this approach from the genetic diseases I've told you about to things like malaria with the Gates Foundation. And so ultimately, if we can be successful in everything I just described to you, we can have an incredible impact. We can fulfill that audacious hope that Ricky has in us. But beyond Ricky, we have the chance to impact tens or hundreds of millions of people. And that's something for which I'm profoundly grateful. And I would be remiss if I did not include this slide. I know Control Labs was up here yesterday, another LuxFact company. Amazing, amazing conference. I love being here in Helsinki. And I have some amazing investors on my side. So for all of you founders who are looking for great investors, please come find me. I'd love to make introductions. With that, I thank you so much for your attention. It's been a pleasure to be here. Thank you.