 Thank you all staying to the last keynote. And do you know how many talks already before this one and this conference? It's over 100. And I have learned a lot from the keynote talks and the technical talks from the demos. So I first really want to say thank you. I learned a lot. The insights, the ideas, the details. And I also want to take this chance to thank LFN and the local organizers for having such a nice event as a spot. But I come for ONS, ONS, but I think I come for something nicer, chocolate. I'm a chocolate collector, so if anyone is interested in chocolate, we definitely can talk after this. It's just so nicer when I add the chocolate to the ONS template and make the 5G AI orchestration look nicer. Cute. And the animals. And the cocos. They're just cute, lovely. So I don't exactly know what I should cover because I'm the last one. And the time when I prepare my slides, I didn't know what other talks would be. So to be safe, so I put all the terms on the screen. 5G AI orchestration. And so talk about 5G. In China mobile, well, all the operators are probably in the world offer 5G available in 2019. This is our offering, 5G to be available, commercially available for 50 cities, major cities into 2019. You might try to look at where the red dots are. You don't need to worry, I guess, all the cities in this room probably as a community. You have visited, and you are visiting, and you will visit. I guess it's all covered. All the major cities will be available. As always, China mobile does. We offer more than we promised. I guess if you go to China, a rural place, you got 5G in this year, and don't be super surprised. And 5G+. And I don't know why in the recent couple of years, we all just not just the tech out, we all just love the sign plus. And back to June this year, the top officers from China mobile announced our 5G plus strategy, 5G plus plan. There are a few calculations, if we are good at math. And it depends on who taught you math. You learn your math from your art teacher or humanity teacher, you get a different answer. So first one is the 5G plus 4G. We have 5G available for a long period of time, and we know we're going to operate the network of 5G plus 4G. And we need to find the best way to have both together and to offer the top quality network for our customer. And the second one is 5G plus 5 letters. They're just not simple letters. It means a serious business. It means where we're going to put our investments on. A, I, C, D, E. A means artificial intelligence. I means an internet of things. Cloud computing, C, big data, and Edge. I don't think we miss anything big. With all those great technologies that are coming up, I think it's important to have 5G or a network to be more innovative. We can do way more innovative with 5G plus these technologies. And each letter means something pretty big. I will not going to cover all of them, but I will open up the A part, the artificial intelligence, what we have done, and what we have in mind. The ecosystem, 5G is not an online business. We're going to empower ourselves. We're going to also support how we can foster an ecosystem to empower other industry sectors. Among them, there are nine major industry sectors that we strategically invested in. I just examples are finance, manufacturing, smart city, transportation, education, agriculture. There are nine of them. And we did a reorg, have organizations, and dedicate to put 5G for those industry sectors. OK, with that big picture talk about, now we come to 5G plus A, artificial intelligence. Last O&S, I talk about a little bit of what we have done for artificial intelligence. But now we announced the platform, the engine called Jiu Tian means the highest heaven or the ninth heaven. It's come from a very well-known Chinese poetry, and in the end, you're going to see how that poetry looks like. So pretty much we're trying to apply AI for five major areas in the business, customer service, marketing, management, security, and network. So we collectively would build the platform to unify our internal AI efforts to be a same platform, provide the computation-powered data, computing, and network environment algorithm solutions. We also use this platform to enable other industries and talk a few applications from the last conference to this one in the past couple of months. A few larger-scale applications that we deliver to the business value, I think is a larger scale to just give a couple of examples. For the management, we apply natural language processing and data analytics into auditing, the invoices, and the paperwork, and our formal contracts. And we also apply the conversational engine on our OA system. And a few other smaller and really fun ones I'm not going to cover here. And for the marketing, we put quite a lot of effort in working on the recommendation technology, or apply that to run our home TV service. That really helps a lot, improve our user experience, improve our revenue. And we also enable our business to serve a few hundreds of hospitals using our conversational agent. And customer service. This one I mentioned last time, we have chatbots. We have the routing, call routing agent. We have the back office, lots of smaller tools put into the customer care back office that really drives on the customer care costs. That's why we can offer really lower data usage rate to our customers. And the harder ones are the network and the security. So we have put out a few applications into network. Here I list one example. Basically, it's using natural language, technology, speech recognition, and the plus machine learning for the decision-make. And when we hear a customer call us to complain a certain problem, we're going to have to dive in to find how we can solve this problem. That takes a long process to hear the message and you transcribe to certain texts. We have to extract certain information from. Then we have to have someone who really, with a lot of network maintenance experience to go into to check into probably over 20 different databases to pull out the data to analyze what was the problem. Basically, we collectively combine these technologies and to automate this process. The efficiency got improved 90% to save quite lots of money for the company. There are a few others and just examples to give here. For the security, now security can be for information security, for network security. Just to give you one example how we help our customer to fight the spamming phone calls. Because nowadays, not only you have a human call you to try to sell you something, you have lots of robots call you. So we put out the technology for using speech recognition, using speaker identification. And based on and plus the data analytics trying to detect which calls are spamming phone calls. That was the only central TV, one of the popular programming in China in back to August. I mentioned we hosted a nationwide competition. And we have two, one for a smart family, the other one for a network. We have close to 400 teams joining the competition and the final, the 40 top performing ones are selected for second round. After this trip, after I go back to China, we will have the final competition. We have two topics. We give the data network data flow prediction, a wireless fault due to cost analytics. Just I know several companies here are involved. Some developers are here also. Thank you for the work. Really, we push ahead the accuracy and the performance and technology. And orchestration, this is a big thing for the community for own up. And what we're going to orchestrate. I mean, what orchestration means for 5G. And we have the NIC network as a slicing. And you can look at the horizontal line and the vertical line. There are way more things that we can orchestrate. You can orchestrate wing F, SDN, network services, slicing, edge, container and run part. And we need to do it across vendors, we need to cross domains across layers. So if orchestration was the option for 5G, I don't think it's option anymore for 5G because the service offer will be way more complicated and we opened up the bottom infrastructure has changed. So I'm thinking orchestration is important. For 5G, it's not an option anymore. No matter the depths and breadth for the orchestration. What are we going to orchestrate? So is life good now? We have 5G and AI. You can orchestrate 5G. You can use AI to further help 5G and we can put all things together. But I want to seriously talk a few questions. I think we still need to work on. When I look at the O&S page of the schedule, we have more than 10 talks on AI or at least the matching AI about different tastes, different flavors, different parts of AI. What is, do we, I just trying to give an overall picture of where we stand on the AI for telecom, for the network. So we look at the standardization organization, so how they think of putting AI and the future network together. More of on the infrastructure level and how AI, the component, should be connected. We talk about AI for edge. But I'm thinking is 5G and AI has, they mutually help each other because there is a challenge for the AI technologies. It's really expensive for any AI technology. And after you developed how you can really deliver a larger scale and you offer that into different industry sectors, you deliver the business value on top of the technology is a challenge. And I think a 5G is here can help to spread out the technology, really make it AI, more UP critters, available everywhere. But should we, so all the devices, 5G is going to connect and will be empowered or enabled about several different levels of intelligence. So now the question, and so it's safe to give questions to you versus I answer all the questions. So I wanna talk about should AI, should our network itself should be intelligent? How it should be intelligent and where we are now. A few questions for discussion. So I mentioned a couple examples. I can see in the literature, not only in industry, in the AI conferences, there are certain dots, certain spots where you apply AI technology for certain parts of the network. My question is, should AI be systematically available for 5G? Do we have the infrastructure or do we seriously think what's the best way we collect the data? Which kind of way we collect the data is most the cost effective and really can reflect directly and cost effectively and to sense the quality, the status of the network to sense the user experience. So I'm thinking, have me lots of thoughts and into the standardization of paperwork and in each companies. I don't think we are, they are yet to, you know, how, of all we should do it. Second is, what is, but when we talk about AI, I just came back from a, before this trip, I attended its international conferences on speech recognition. So I always think we work on network intelligence but what is our main math problem? I think I mentioned this last time is on what level we need to do the abstraction. So we can, we can take the AI technologies which mainly developed for speech, for image, for natural language processing will apply to our network for analytics and is that enough? Is our problem unique enough to need different set of technologies to solve and what we can contribute back? I don't think it's clear but I'm thinking the AI technology is developing, it's improving at a pace which I have never experienced in the past 20 years in the AI field. So I just want to add a few words. So for the, for the, for the mainly for the AI with the, the current state of art AI technology pretty much designed for image processing, for video, for speech, for natural language but the network is different kind of a problem personally I think because we, the AI technology is trying to, trying to do what we human unconsciously to do because we recognize face, we don't know how to describe it, how we recognize face, we recognize voice but we don't know how to do it and then you design a technology you are trying to do what we unconsciously, how we unconsciously think but our tech, our network is what we, is human designed, is not a naturally formed thing. So, but the network become way and way more complicated and I'm thinking AI coming instead of handling the unconscious thinking here is the needs to handle the complexity. So talk about where the intelligence come from or when we say we want our network to be intelligent to be ready for so we can, so the network can connect all the intelligent devices so the intelligence there's three different resources where the intelligence can come from you can learn your technology you can learn a model from data you can build up on top of your knowledge base you can, you can have a environment where you can do trial to get the feedback and then you can find your, find cue your model but thank you for network we are lack of for the three foundations I don't think we are anywhere closer we don't have a huge corpus which can represent our typical problems in the largest scale available open and we don't have that huge knowledge base we can all leverage on to manage to operate our network in an intelligent way we don't have an environment we can do trial on everything's pretty small if you look at the other technical papers even in the top conference for telecom and all the professors are trying to have a very small network in their lab and then they do the experiments on which is far different from what we see in the real industry and what are the state-of-art AI technologies you know we often think AI is something you give data you give it X and try to generate to Y but AI has been quite different from what was before we needed to really know and extend to the technology then know how to put that into the network operation is our problem different from the typical AI problem as of today I already mentioned a couple of ideas we can have more discussions so where we stand we, you know, I think a China mobile made a decision we determined to work on solving these fundamental science problems and engineering problems with the industry and with the researchers in the field we established a few labs with the top universities to collaborate we look forward to collaborate with the community too so at the end I said the name or the engine of our platform Jiu Tian come from a poetry so it looks beautiful it's hard to translate so we'll leave it that way and last page and thank you all and we have a few booths touched a little bit what I talked about if you haven't come to the demo or to the booth please do, thank you