View transcript
How can you positively impact the lives of a billion people within a decade? That was the question we got, and that is the reason why we started our company, Iris. My name's Anita. I'm the CEO and co-founder. I have three other co-founders in this venture. And what we are doing is that we're building an artificial intelligence that will read all of the world's scientific research and help us connect the dots in it. I was told not to pitch, so I will not do that. What I want to do is give you kind of the bigger picture of what we're trying to achieve. I want to talk a little bit about our moonshot, a little bit of this science fiction novel that we're writing about our company, not in the literal sense of the word, but also how we combine that with kind of the nitty-gritty details of actually running a company, getting product market fit, and getting the real stuff going. The problem that we're trying to solve is that every single day more than 3,000 papers are published in science, technology, and medicine alone, not to talk about the millions and millions of papers that are out there. And we believe that if you can connect the dots in all of that body of knowledge, that we would solve a lot more problems. There's so much information. We have so much knowledge. And that's what we're trying to do. Our human brains can't read and understand everything. So if we build an artificial brain that can do it, essentially we should be able to make sense of that data. Now that's kind of a 10-year undertaking or so. It will take us about a decade to get there. But we decided that we wanted to launch a product within three months. So that was kind of our mission once we started. We want to get out there, we want to do the lean approach to it, get something out quickly, test it, fail fast. So I'll take you through a little bit of our journey, what we have been doing and then what we will be doing. And I'll mention that we started up in August 2015. So we're a fairly young company with a very big ambition. We decided to start by mapping out scientific research. And we decided to start with TED Talk. So our first product, which is live on our website, is essentially a tool to show you the science around a TED Talk. Sort of engaging, kind of fun, and a good place to start. It's also a really nice kind of body of text to start with. And we're using transcripts. We're not dealing with audio or video. So what you do is you take the link to any of the main TED Talks on the TED site, drop it into Iris, we extract the key concepts, map them out, and let you navigate around and find relevant research papers on each of the different topics touched upon in the TED Talk. And that's where we started. So we got this out within three months. And to get it out there, we have about 10,000 visitors a month, and it landed us our first sale. Our Future Health, a e-health conference in the Netherlands, wanted to use our tool for their conference in order to map out the science around their talks. So we did our first sale, which came a little bit from the left side. We didn't quite plan for it, but it was a great first early revenue. But this, of course, is fun and engaging, and we can showcase and demo a little bit about what we want to do. But we want our next version to be something that people will actually use in their daily lives. So what we are building over the next few months, and we're launching the first version of this in September and then over the next couple of years, basically for any innovator, R&D department, researcher, sorry for the dark colors, you can take any scientific text, over 500 words, drop it into Iris, and we map out the scientific landscape around it, including giving you some trends and statistics on where the research is happening and what's hot, so to speak, in the research world around this. We are working very closely with pilot customers to actually make this happen, to make sure that what we build already now will provide value and will provide a tool we can actually make money from in the early stages as well. But let's do a little bit of a jump. In 2020, we will be able to not just map out scientific research, but actually start, Iris will be able to start responding to specific prompts. So if you have a question you want to solve and you need a specific method to that scientific query or the invention you're trying, we will be able to break down the text and help you find that method. Let's say you're working on a specific protein and you want a method to, you know, whatever you want to do with that protein, we can help you find methods from across research fields. So you could break it down, and essentially, Iris will be that science assistant that helps you make sense of the science around your project. Now, if you move 10 years into the future, Iris will not just map out the research and help you understand your project, she will also be able to do her own projects, right? So we can feed her a body of content, let's say related to climate change, Iris can deep dive into it, make sense of it, come up with a hypothesis, test, validate, even publish a research paper on it. Now, once we've got that down, Iris will be able not just to, you know, train human beings and herself in scientific knowledge, she will also be able to train other artificial intelligences. So let's say a swarm of drones that will be implemented in a new region that has weather conditions that that system is not set up to deal with, Iris can help them learn about weather conditions in that region as a simple example. As I said, our vision is kind of a moonshot, that is where we're working towards, and the area of artificial intelligence, I know it's kind of a hype right now, and there's a lot going on, and a lot of people calling themselves artificial intelligences or calling their companies that. I mean, we are very aware of what we have right now is not an artificial, a full artificial intelligence, but we believe that we can get there. So what we're doing right now is we're trying to get to the point where what we're doing right now is the way we go about it is that we started, the first demo version, the first version we have out, we started that with kind of code and ways of working with it and technology that was already well developed, 10 years old, 10 years plus, just to get it working, right? Then we can quickly move up to kind of state-of-the-art technology, the algorithm that we're implementing right now was proposed in 2015, and that's just within six months of the previous launch. Then we can move up to kind of cutting-edge technology, and eventually, as our CTO says, then it's our time to shine, and we can start building things that haven't been built before. More specifically, we want to work with, so what we have right now, sorry if you can't see this, we're working with all unsupervised learning, we're adding in a layer of supervised learning in a neural net, that's what we're working on right now, and then eventually, come the next couple of years, we're going to add in reinforcement learning as well for text, which hasn't really been solved yet, and that's where we believe that our technology actually will push the field forward. Finally, and I'm going to wrap this up and see if there's any questions, but I gave more of a regular pitch at a 500 Startups event a couple of months ago, a couple of months back, and one of the investors, I spoke to him afterwards, he only had one note about my pitch, and that was either, this is genius or she's really good at bullshitting, and I hope it's the first, that's what we're working towards, we believe in what we do, but there are some specific challenges with dealing with this kind of moonshot company, right? We're aiming really, really high. One thing, for example, when we talk about our marketplace, how much can you charge for an educational license for an artificial intelligence, as in AIs training AIs? What does that market look like? It's a complete blue ocean. Now, we know our first baby steps, and we know that market, we can charge $500, there's 300 user licenses per company. We know a little bit of things like that, but that's not a really, I mean, it's a nice scalable market, but it's not really a unicorn kind of market, but this blue ocean of AIs training AIs is kind of cool, but how do you put that into numbers, right? Same when we talk to our corporate users and our pilot users, it's always about managing the expectations of what we actually will be able to deliver in September versus what we can do with their help together with them in a year or two or five, and it's always this balance of not overselling what we're able to do, but actually still solving real problems and finding that problem that can be solved with the technology that we're currently employing. Finally, we are a Norwegian, tiny little startup company, although we are fairly international already, all my co-founders are not Norwegian. We looked at the investor landscape, Norway, I mean, we have a way to go, but I am also very positively surprised that we've found, we're currently about to wrap up our pre-seed round in Norway, and we have found a small number of angels that have been willing to kind of take their bets with us, go in and believe in our big vision and that we can pull it off. So I'm very positively surprised about that at the end of the day. So I want to end with a quote, a friend of mine, David Roberts, he says that, you know, running a startup company is hard, whether you aim low or aim high. So you might as well aim really high, right, because it's going to be hard either way. Thank you.
