 Buenos días. Muchas gracias por venir. Un placer estar aquí con todos ustedes. Hi, my name is David Gorena. I'm a principal program manager at Microsoft and part of the artificial intelligence organization and today we're gonna be talking about AI and knowledge from a Microsoft perspective. So I want to start with a story because I like stories and the point of the story is to tell you how technology infused by artificial intelligence has dramatically changed and helped my life over the last year. So a year ago I decided to move from Mexico to the United States. I accepted an offer from a company like Microsoft and it changed my life, it changed my family's life as well. So one of the things that was happening at the time was that my father had been diagnosed with cancer. So it was obviously a difficult decision, but I knew that I was moving to the US, has some of the best hospitals in the world, so it was an opportunity. And I started learning that there are machine learning algorithms that are able to detect and exploit weak links in cancer that doctors can then go on target to try to get rid of it. Pretty amazing. Another thing that was happening is that I knew little English. I was used to speaking Spanish all the time. So moving into a country where people speak English was tough and even tougher to go into meetings not knowing the context of things and trying to figure out what was going on. Again, I learned that there was technology that could help translate the documents that I was seeing during the meetings and not only that, but the presenter's voice could get translated for me in real time by captions so that dramatically helped me improve my English skills and also understand the context of all these meetings that were happening. Another thing that was happening, well, I obviously was moving to a completely new city for me. I knew little to nothing about Seattle and that was tough, but I started learning that there was technology at my hands to help me move in a city that I knew nothing about. So, for example, I was moving in downtown and I saw a big stadium. I could either search for the stadium or just take a picture of it and I would see data about it. So, I would learn that I was looking at Century Link Field. That's where the Sounders and the Seahawks play, but not only that. With time, I started seeing that technology would learn from me and it was able to give me recommendations. So, for example, I noticed that technology started learning that I liked soccer and American football and it started recommending me games for the Seahawks and the Sounders and actions related to them. So, click here to buy a ticket for the next game, which is happening next Sunday, for instance. So, again, big things happening and technology right there helping me. I was moving my family as well to a completely new neighborhood. I didn't know if the neighborhood was safe or not. So, one of the things that I immediately did was buy a home security system and I started learning that there's technology that can identify people in your house. It learns over time, it learns the people that are not part of your family and when it detects someone that shouldn't be in your house, you get a text message and police are called immediately. So, again, technology at my service helping me feel confident, feel safe. I could go to work knowing that my family would be all right. Now, this sounds like a very successful story, very cool story. But usually stories aren't that good and there are a few gaps in the story that I just told you. My journey in the U.S. didn't start one year ago. It actually started 15 years ago. So, a lot of the technology that I was talking about didn't exist back then. So, in the healthcare area, my dad was diagnosed with cancer but by the time they detected it, it was too late. Moving to a new city was very tough. I didn't have this technology that exists today. With Bing and the Bing applications, you can take pictures, you'll get data of everything you're seeing, the application that technology will learn about you and will start giving recommendations that didn't use to exist. So, for me, 15 years ago, it was hitting walls, hitting rocks, learning about a new culture, learning about a new city. And Bing in meetings was probably one of the toughest things because it is true. I knew little English, so there was a barrier in the language itself but then on top of that, understanding the context and the technology words that were being used by people, that was even harder. Well, today, the technology that I described exists. Microsoft PowerPoint is an example of it. It not only translates the documents, the decks that you're seeing in the language of your choice, but it actually translates what people are saying in a presentation in real time through captions, as I mentioned. The home security system one, that one is a tough one. I'm still looking for something that exists like that. I know it's possible because the technology is there, but we're still not there. That's the truth. And actually, I'll tell you another story. Around four months ago, I was woken up in the middle of the night by my home security system. I was sleeping in Seattle and when that happens, well, first of all, you're sleeping, so you freak out. And you have children sleeping. I have my family living with me. So it was a freaky moment. And the first thing that I did was grab my baseball bat and I started walking slowly down the stairs, hoping that there was nothing in there, and myself on ready to call the police. So it ended up being that my alarm had been triggered by a balloon, a balloon that my daughter had left in the living room the night before. So this is a great example of how technology can be infused with artificial intelligence to make our lives just simpler. And the technology is there. We just have to implement it and put it at our service. So the point is that artificial intelligence is being infused everywhere, in our work life, in our play life, and everything we're doing. It's happening. It's been happening for years. Just like the example that I told you, sometimes we don't realize that this big change and that this revolution is happening, but it is every day. Now, one of the things that I want to mention about artificial intelligence is that there's at least three things that need to happen for a device or an application to be AI capable. The first one is that the device or the application needs to be able to understand ambiguous data and learn from it. So going back to the home security system example, this would mean that my home security system can understand that I go to sleep every night, and that I need it to be active all the time. My home security system should be able to learn that every morning around 7 a.m., I go downstairs for a glass of water, and it's me. I'm supposed to be at my home, so the home security system should be able to understand these things. The second thing is that they need to be able to interpret the environment. So, again, in the example of the home security system, if I'm talking to my wife in the living room, my home security system shouldn't do anything about it. But if I'm talking to the home security system, it should be able to know that I'm talking to it, interpret my voice, change it into text that its internal algorithms can use to be at my service. And the third one is that these devices or applications should be able to lower the barrier between human beings and machines and applications or devices. And we're going to be talking about this further in my presentation. So, let's take a step back and try to define, in a very simplistic way, what is artificial intelligence? So, the way that technology has worked for many years in the past is that human beings would learn about the technology, become experts at it, and then go and teach other people about it. That's changing. Now, all these devices and all this technology is learning about us and are helping us not just in a reactive way, but in a proactive. So, it's basically amplifying our senses to be able to anticipate, help us anticipate, what is happening. So, for example, if I'm driving a self-autonomous car and I, as a human being, don't see that there's a stop sign, there's AI right there that can see the stop sign for me and help stop the car to prevent an accident. So, that's right there, technology learning about the environment, learning about the situation, learning about the situation to help simplify lives. For people that like numbers, 95% of customer interactions are going to be powered by bots by year 2025. 75% of apps will have some form of artificial intelligence by the end of this year, and we can see that already happening. 30% of search results will be based on voice by year 2020. And I like the last one, which is 20 billion devices connected to the Internet by year 2020. And think about it, it makes sense. I'm a single person in this room, and I have a cell phone, a smart watch, and a laptop all connected to the Internet. So, three devices on one person. And if you go to younger people, you can see much more devices on a single person. And this is very important as we go into the presentation because we're going to see the 20 billion devices out there. There's so much data happening and so many possibilities of AI learning about this data to make devices and applications intelligence and be at our service. So, why now, though? Why are we talking about artificial intelligence now and not 10, 20, 50 years ago? If we've had calculators, powerful calculators out there for so many years, the technology that makes cars drive by themselves have been prototyped since many years ago. What makes this time so special? Well, there's a few things, and they come together, basically. The first one, we talked about 20 billion devices by year 2020. So, there's a lot of data happening, and models are training on this data every day to make big things happen. So, that's the first one. Big data that 20 years ago was not there. The second one is powerful algorithms that can take this big data and do something with it. These powerful algorithms that exist today didn't exist 10 or 15 years ago. And the third one, cloud computing. So, you have to have somewhere to host these applications and these huge amounts of data. So, cloud computing is a reality today. It is scalable, and it is cheaper than it was 10 or 15 years ago. So, you can see how these three things combined basically make it the ideal environment for artificial intelligence to be striving right now. And the good thing is that Microsoft has been thinking about all these things for the last few years. And it has a portfolio that I'm going to be talking about today. At a high level, the portfolio contains an agent that knows about you and that learns about you. Applications that we're infusing with AI every day. Services, so if you're a developer and you want to be able to infuse your devices or your software with intelligence, Microsoft provides these to you. We're going to be talking about them. And then the last one, infrastructure. As I mentioned, cloud computing is a requirement, needs to be there, and Microsoft has an offering for this. And all these components sitting on top of a very important layer of knowledge. And if you were part of the keynote earlier this morning, Oscar talked about this and he basically said it and I think he made my job easier for those of you that went because he basically said that AI cannot exist without meaningful data, without knowledge. And we're going to be talking about why and how today as well. And actually my next few slides are going to be about this, about knowledge. At Microsoft, my team is a team that owns the knowledge graph. So I've been doing this for the last few years. So what's knowledge? How do we represent knowledge at Microsoft? Well, we have a knowledge graph. Think about this knowledge as anything that you can think of in the physical world can be represented as an entity. So for example, I'm a person, I'm here, and I'm an entity. And I have attributes, I have properties. So I have an age, I have a height, I have color of my eyes. Those are all attributes or properties of myself. And then there's other entities that I interact with every day. So for instance, I like a certain song. So that song is an entity. I watched the movie last weekend, that movie is an entity as well. And you can see how entities start relating to each other. And knowledge has been powering search engines for many years. The easiest example that I can think of, if you go to bing.com today and search for Madrid, you're going to get an entity card. That entity card has Madrid with its attributes. So examples of the attributes would be the description of Madrid, the local time, the local temperature, or whatnot. Because we are in a graph, we know things that Madrid is related to. For example, we can tell you that people that have been searching for Madrid have also been searching for other cities. So those kind of things and those kinds of relationships exist just because we are in a graph. Now why is knowledge important in powering AI? Here's an example. And let's go back to the home security system. Example, imagine that my home security system is AI capable and I can talk to it. I can tell it, hey, home security system, I have a cat. Whenever you see my cat in my living room, don't trigger. That's fine. He's part of my family. So translating my voice command into text that algorithms can use, that's what AI can do. Easy. But if the home security system doesn't understand what a cat is, and more importantly, if it's not able to differentiate that cat from the millions of other cats out there, then AI will not make sense. So that's knowledge right there helping AI make sense. And there's challenges to this graph, having this graph. The first one is coverage, what we call coverage. So the knowledge graph has to know about the world, has to know about cats, animals, other human beings, songs, music, flights, has to know about everything. Otherwise it's not going to make sense. This data needs to be fresh. So flights are being canceled or delayed every second. People are being born, people die. If the graph is not able to cope with these changes, then the data in the graph is not going to make sense. And there's the correctness aspect of it. So if someone goes to Wikipedia today and starts writing about myself, about David Gorena, and he says that David Gorena was born in Australia, that's incorrect data right there. How does the graph know that that particular data from David Gorena is not right? That's something that the knowledge graph has to be able to cope with. And a quick example of some of the challenges of a knowledge graph, if you go to bing.com today and search for Will Smith, because of popularity signals that we get, you're probably going to get this guy who is a famous actor in the US. Now this entity card was created from 41 different sources. Imagine 41 different sources telling different things about Will Smith and making sure that we grab the right data, the most fresh data. If you go to Wikipedia today and search for Will Smith, you're going to see that there are hundreds of them. So being able to take all these data and conflate it into different entities is one of the big challenges of a knowledge graph. So at a very high level and very simplistic way of describing how to create this knowledge graph, I would say, number one, you have to be able to ingest this data into the graph. And there's two ways to do it. One is from structured data. So think about this as templatized data that is easier to ingest. But there's some structured data. So there are paragraphs of data out there in the web that you can grab and try to identify entities from the paragraphs and then conflate it into the entities in the graph. And that's the second step, being able to conflate all these sources into the different entities. Then as a third step, you start creating inferences about how all these entities interact with each other. And finally, you publish this data into applications. All right. So we talked about knowledge and why this is very important. Now let's go through the different components in the portfolio. And we'll start with agent. One of the agents at Microsoft, we call her Cortana. As I mentioned earlier in the presentation, one of the very important things about AI is that it needs to be able to understand ambiguous data and learn from it. So your experience with Cortana, or an agent in general, in day zero, is going to be very different from your experience in day 300. Because if you allow this agent to learn from you, the experience is obviously going to improve. So Cortana knows about your life, knows about your work, knows about the world as well. So I have a business trip coming up to Madrid. I should be able to tell Cortana, hey Cortana, I'm going to go on a business trip, make sure that my home security system is on while I'm out. That should be something straightforward. Cortana knows about me, knows about my schedule, and knows about my devices. Now Cortana, as I mentioned, can learn from you. It can learn from your calendar. It can learn from your emails. So if Cortana starts seeing that every time I go on a business trip, I email my neighbors about it, letting them know. Then Cortana is going to be able to learn that pattern. And now that a trip is coming up, she can send me a reminder, or even better, ask me if she wants to just let my neighbors know. And then again, Cortana knows about my calendar. So if I ask her about my trips, she's going to be able to give me a card with all my trip details. And obviously very important is that this data needs to be fresh, because trips are delayed, canceled, or whatnot. And Cortana also knows about people. And as you probably know, LinkedIn was acquired by Microsoft, so Cortana knows about LinkedIn people as well. So one of the cool things that Cortana can do, if she sees that I have an upcoming trip to Madrid, she can tell me that people that she knows are going to be at the conference, that I know, and she can give me more details about them. And not only that, she can give me information about emails that we have interacted with in the past, or she can tell me about meetings that I have had with them or are scheduled to have in the future. All right, so that's Cortana. Now, as part of the Knowledge Graph team at Microsoft, one of the things that I do on a day-to-day basis is infuse this knowledge and AI capabilities from the platform into applications, specifically Office. So even though Microsoft is working on infusing this AI into all applications, today I'm going to scope it to Office applications. I'm going to be talking about two of them. The first one is Microsoft Search. This is a product that was announced a few weeks ago at Ignite in Orlando, Florida. And the idea behind this product is that we want to take Search at Microsoft and put it everywhere, including obviously the Office applications, and make it consistent and coherent for you. Now, this search is going to be changing. If you use Bing.com, you can see that intelligence is already there. We want to make sure that this happens across applications. So it's not just a reactive search, but a proactive one. So based on the context of whatever you're doing, the search results are going to appear based on this context. So if I'm in a PowerPoint presentation and I'm interacting with certain people, the search could guide me into very quickly sharing my document with these people or asking them questions or whatnot. So consistent, coherent, and more intelligent search across Microsoft products. This second one was also announced at Ignite, and the rollout to production actually started last week. It's called New Data Types in Excel. The idea is this. Anything that you can type into Excel today is text. The idea is that with a click of a button, that text can be converted into rich entities coming from the knowledge graph that make your experience much more powerful. For this V1 that we started rolling out, we're focusing on two segments, geography and finance. So for geography, think about things like countries, states, provinces, etc. And for finance, you can think about stocks, mutual funds, etc. So let's have an example. We've convinced the Spanish government to go ahead and install home security systems everywhere. How can we prioritize this task? One of the things that we could potentially do is get the list of Spanish provinces into Excel. This is going to be text. Now we select the text, we hit the geography button, and this is not text anymore. They're rich entities. So now you can pull information from the knowledge graph out of these entities very easy. So you can get things like population and area, and then using the calculations that have existed in Excel for years, you can quickly calculate population density, and then this could be a way that you prioritize where do we go and install home security systems first. Just like for geography, oh, and by the way, if you hover over these entities once you've converted them, you'll see entity cards like the ones that you're seeing in the screen. And the same for financial instruments. You can have fresh data in your Excel spreadsheets pretty much all the time now. And these features are not only about data, they are about intelligence. So if you had two lists of places, and you had China in one of them with other countries, US, UK, et cetera, you convert them, and the China is going to be converted into the China country. But if you have China in a list of other Mexican municipalities, you're going to get that China converted into the Mexican municipality because it turns out there's a China municipality in Mexico. So not only about the data, but about the context, which is what we're trying to do here. All right, so the third component in the portfolio is services. So if you're a developer and you want to be able to infuse this kind of intelligence into your own products, that's available today. Today I'm going to be focusing on two of the services that we provide, the first one is cognitive services, and the second one is the bot framework. So for cognitive services, these are basically APIs that you can use to infuse this intelligence into your products. They're easy to use, they're flexible, you can write in the platform of your choice, and these have been thoroughly tested by Microsoft for the last few years. So what are the kinds of APIs that we provide? This right here in your screen, let's take the example of the home security system to go through this vision, which is the first one. If you want your home security system to be able to differentiate between a balloon and a human being, these are the kinds of APIs that you would be using. If you want your home security system to understand your voice commands, then you would be using the speech API. If you want the home security system to be able to understand not only English, but also Spanish, French, German, then you would look at the language API. And then the last two, knowledge and search, they're very tied together. These are the APIs that basically give you the knowledge of the world that comes from the Microsoft Graph. So if you want your home security system to understand what a cat is, what a window, what your walls in the house are, then you would be using these kind of APIs. And this is the Microsoft bot framework. So we talked about it in the numbers. It's going to be 75% of customer interactions supported by bots. So this is very important, will become very important. And as Oscar mentioned in the keynote, it's not about having stupid bots. It's about having bots that understand about the world, that understand about knowledge, so that they make sense. So the framework provided by Microsoft supports this knowledge and is scalable. So you write your code once and you're able to deploy to the Microsoft channels and third-party channels as well. So going back to the home security example, imagine, as we had mentioned, we convinced the Spanish government to go and install these security systems everywhere. Well, they'll probably want to have a website up for people to be able to go and ask questions. The way we would do it a few years ago, we would have a Q&A page where people would go and search for a question, find the answer. Well, this can very easily be changed to be a bot. The bot knows about the Q&A list. You shoot a question to it. The bot is able to just speed the answer. And this is a very simple scenario. We can get into more complex stuff. You can have a bot help schedule the installation of the home security systems in order to be able to do this. The bot needs to know about the world, needs to know about people, needs to know about their houses, how many sensors they need to install in their houses, needs to be able to have a smart conversation with people, basically. And this can be done today. All right. And the last component, which I'm not going to drill into today, but as we have mentioned, it is required for AI. You need a place to host your algorithms. You need a place to host all this big data that you're going to be training your models on. Microsoft provides a solution through Azure. So I want to summarize what I've said here today as AI is already changing the way we live, work, and play. And if you think about it, this has already been happening for the last few years. As I mentioned in some of the examples, AI is already helping in the healthcare industry. AI is going to be able to save lives because it's not just a reactive, it is proactive. It's learning about what's happening, finding patterns, finding trends that will help people live better lives. It's going to help us live more simple lives. We go to an unknown place, we can take a picture of something, and we get all the details. And technology starts learning about us, and suddenly it can start recommending stuff based on a personal use of technology. So this is a revolution that has begun. You can either be part of it or completely lose the train. My advice to everyone, my homework to everyone is to go and read what's out there. There's a ton of things happening, not just at Microsoft, but in other big and small corporations. People are investing a lot of money into this. This is going to be a trend over the next few years. And at the end of the day, this is about helping people. It's about empowering people and making their lives simpler. So if you're a developer, go and take a look at the tools and the APIs that are out there. If you're not, go and play out with the products that Microsoft and other companies are putting out there to help simplify your lives. And please give us feedback. Muchas gracias a todos. Hello, good morning. I was wondering if you had any plans of incorporating artificial intelligence in Microsoft Power BI, if you know about it. Yes, so very good question. The question is, are there plans to incorporate artificial intelligence into Power BI? Yes, it's definitely one of the things that's happening now. I can tell you that as part of working with Office every day in Excel, we're starting to look at Power BI because basically the kind of things that you can do in Excel in Power BI, you can scale tremendously. So yes, it's in the plans. It's already being worked on. I can't talk about specific details at this point, but one of the plans that Microsoft has is basically infuse these kinds of things that we're talking about into every product and Power BI is one of them. Power BI, we have the same problem. Cordoba, it's in Argentina and not in Spain. We suffer a lot of these bad understanding of where is the cities, especially when you work in Spanish. When you do it in English for Excel, for the tutorials, everything works fantastic. But for other non-English language, it doesn't happen very good. But my question was related, what about security and confidentiality with AI? For example, even in your house, in your example, you want that Cortana is listening, all the conversation, everything. Where do you put the barrier between? Yeah, so that's a very good question. Privacy, security. So one of the things that Microsoft is proud of is that it designs everything that we make with privacy in mind and trust, the trust that we have with customers is the most important thing. So whenever we design a product like the one that I showed in Excel, we do it from the beginning with this privacy in mind, making sure that your data is in a compliant environment, won't get out of there. And we always have backup plans in case the worst happens. But I think that the summary of the answer is you have to put trust, you have to put privacy at the very beginning in the designing of these products to make sure that the data is stored securely and that you as a customer can decide if you want to share some of this information for the models to be trained on. So keeping your data safe and then you as a customer deciding if you want to share and at what level your information. I think those are the two important parts of it. Is there any way to filter, for example at Cortana, if we are going to discuss confidential things, how do you turn it off? Yeah, I mean there are settings in Cortana today. You can go to the Cortana settings and say the data that I'm going to be using from now on, that I'm going to be speaking, I want Cortana to not know anything about it. There are settings today where you can do that. Thank you. Other questions? All right. Thank you. Muchas gracias.