 Hello to everyone out there in open networking world. We really wish that we were able to be together in person, but we are still living in this world of taking our health seriously. We have what is a super session today. We have both an incredibly amazing and interesting presentation from a number of experts within China Mobile, followed by a panel with experts and executives from China Mobile and Verizon. You are getting basically a double dose of expertise from some of the most important and significant operators in the world in this session. So I hope that you are prepared to take this ride because it's going to be a great one. So as I said, that is the structure of our session today and we are going to start with a presentation coming from four experts at China Mobile. We have Leihuang, Yan Yang, Yuhan Zhang and Casey Liu. All of him are incredibly well-versed scholars and researchers who are doing the work to transform the largest mobile operator on our planet. And so with that, I would like to hand it over to Leihuang to start the presentation. Hello, everyone. Welcome to the session, the state of telecom operators adoption of intelligent networking. As the most popular technical field in recent years, AI has been widely used and researched in communication industry network management, operation and maintenance are traditional communication industry services. Operators and vendors have done a lot of exploration on how to make network management more automatic. However, automation is not enough under the current technology trend. We should make the network intelligent and autonomous. So today we have scheduled this session for sharing the state of telecom operators adoption of intelligent networking. This session consists of two decks, presentation deck and panel deck. Let's start with the presentation deck. In the presentation, we will firstly introduce what is intelligent networking, then share with you about what operators have done in trying to build intelligent networking ecosystem, include ELAG intelligent networking survey, white paper and some other current works across community organization, and we'll introduce some specific best practice from operators. In the end, we provide recommendations for reference. Last of work we did in the early stage was networking. So what exactly is network intelligence? Combining the current definition of network intelligence by various organizations will give a reference explanation that is network empowered by AI technologies and systematic integration of AI and communication network on hardware, software systems and processes to realize lower cost, higher efficiency and agile business. And currently there are various open source and standards organizations that have contributed their efforts in the process of exploring the realization of network intelligence. In order to deeper understand current status of network intelligence in industry, the Elephant ELAG organized a large scale survey in Elephant at the beginning of the year. This is in this topic, we will firstly introduce the survey and analysis combined with ELAG current work. We will start out and share ECSP's requirements and explorations in AI and intelligent networking. You have as the main contributor will share with us both the survey and the follow-up white paper. Let's welcome Rehan. Okay, thank you Le. Today I'm glad to have this opportunity to present our work on intelligent networking and AI survey and white paper. Please down please. The intelligent networking survey is DOH's first survey that opened it to the NPR and AFRN community. A total of 65 people participated and those participants covered such a wide grant. We consider that survey result data could basically represent the requirements of AI industry. Please down. The survey contains these six parts. Based on the feedback from survey, we conducted an in-depth analysis, but consider the season time is limited. We only take a brief introduction about some key findings from the survey feedback. Please down please. The first key finding is about intelligent networking's current progress in industry. As you can see from this page, nearly half of operators are still in early stage opposite O&M, and only about 16% of vendors have a long-term plan for scale development of AI into their architecture of products. In general telecom industry is still in early stages of intelligent transformation with relatively low-level autonomous networks. Please down please. The second key finding is about strategy for improving intelligent networking. We could find that leading operators and vendors would use an account in its intelligent network application through unified AI platforms. A consensus could be reached that developing unified platform is essential to improve network autonomy. Please down please. The third key finding is about application scenarios, top three AI application scenarios are the industry most interested in terms of 2B operations and maintenance, service assurance and network optimization. AI has been put highest expectations on those scenarios since they are those one directly serving customers and frontline operators. Please down please. In terms of ecology strategy a conclusion could be reached that industry are currently in need of both specification and open certification of intelligent network service and network AI algorithms. Please down please. Thank you. In terms of the engagement of different ICOs and OISs ORAN, CENGPP and ITI-UT are the most popular ones among the respondents engaged in nearly half of the respondents are engaged in ONAP and ORAN. Please down please. And the the bow is a brief introduction for the survey and analysis from the results of the survey. We can get the current status of the industry and the development demands on the industry. Based on the survey analysis results and the SESP requirements the UAG is publishing intelligent networking and ML white paper to identify and highlight the last state thinking and commendations for building and supporting intelligent networking. In particular we first introduce motivation and definition and they are conducted detailed analysis and non-meaning on the network intelligence survey by some results of the survey analysis combined with the current operator work priorities we have collected in UAG and we put forward suggestions for the development of the network intelligence industry from the operator's perspective. Finally, the suggestions and the requirements are implemented to put forward a call to action for construction of the intelligent network because it's that's all. Thank you. Thank you for the good introduction and this white paper as you can see in the screen it will be published in October UAG will successfully successfully hold various meeting form from October to November to promote this white paper include our one summits and in the follow up we will also launch a series of webinars with Elephant to promote the white paper. Welcome to download and read this white paper. We believe you will learn about the development trends of intelligent networking you care about. The next work we have done is UAG data and model sharing project. The reason why we propose this project is that data standardization share data sets and models at a long-term challenges of intelligent transformation that operators must deal with and basic data algorithm framework is also the most required capabilities provided by unified intelligent network platforms. In order to help bridge the work between operators and manufacturers and enable operators and vendors to share data sets and models we have established this project and we work with tech to promote this project to collect network intelligence scenarios from operators and vendors. One of the project we are currently implementing with Anocate is AI ML for telecom cloud use cases. From the feedback of the survey, the highest priority to work on is testing and certification services for intelligent network. Today I have invited Yan Yang who is the CVC Vice Chair to share with us the collaboration work between CVC and us. Welcome Yan Yang. Thank you. Next we will give a brief introduction of the testing and certification for network intelligence from EOAG survey, lack of quantitative indicators and certification service is a problem that must solve from partial autonomy to high-level autonomy. So beauty and open and effectiveness evaluation system has become high priority industry demand. So in order to promote this work, CVC and EOAG are working together to collecting their testing and certification requirements and plan to launch the new branch of Anocate Assurance program to accelerate the development of next generation AAP badges for intelligent networking focus on supporting the telecoms adoption and deployment of network intelligent technologies and exploring their best practice from the full industry. So if you are interested in this work, welcome to John. Thanks. Thank you Yan. After the introduction of the current EOAG work in building an intelligent network ecosystem, we hope to share with you the current operators' exploration and practice through some typical operator practice cases and hope to bring you source and resonance. In this part, I have invited China Mobile's representative Kai Xi to bring us the current practice of China Mobile in the process of exploring and developing network intelligence. Welcome, Kai Xi. I'll take it from here. In this chapter with great pleasure, I'll share with you some of our best practices in exploring network intelligence. Next page, please. Hello. In July this year, China Mobile has issued the first best practice white paper. We have officially set the target of reaching L4 level autonomous network by 2025 with the most complex and large-scale network in China and, in fact, on this planet in terms of number of users, we spend $140 billion on network operation and maintenance every year. So this project is supposed to save us a huge cost and offer better service assurance, deliver automated user experience at the same time. Next page, please. So to achieve this goal, we have developed the closed loop evolution path. Following the top design, every round of evolution starts from level evaluation to identify the common shortcomings. This way, coordinated planning and targeted measures could then be made to tackle the problem and progress could then be made. These will be first implemented in selected subnet to verify the feasibility and effectiveness value. High-value use cases will then be widely adopted. So far, we have successfully launched almost 80 pilot use cases in more than 27 provincial companies of China mainland, covering whole life cycle of network planning, deployment, maintenance, optimization and operation. Next please. Here we have chosen 7 typical use cases successfully deployed and well verified in provincial companies highlighting the business values or quantitative results created by each one. Taking the last one example, the intelligent IDC refrigeration with this use case adopted on the health network, estimated 220 million kilowatts could be saved each year. This will not only achieve lower cost for operators, but also a great step to promote sustainable development and low carbon production. And that's all for my part. Thank you. Thank you, Kai Xi for the good sharing of China Boba House best practice. Based on the current status of our industry practice and combined with the several results of this survey, we think that telecom industry is still in early stage of intelligent transformation with relatively low level autonomous networks in terms of which EOAG will continue to focus on building intelligent networking ecosystem through collect operators, AI and our requirements, collaborate with and promote relevant resource open source communities and project. And consider from specific perspective we have write down some of the recommendations here for your reference and we will talk about it in our panel deck. About the horror content of the slides, thank you everyone for hearing. And thank you everyone of you presenters. This was a great deal of amazing information. And now we transition into the panel portion of this one summit super session on intelligent networking. Now we turn to our panel and I'd like to introduce our panelists. We have Lingli Deng from China Mobile, Stephen Casey from Verizon and Beth Cohen also sometimes known as Machete Cohen from also from Verizon. Thank you panelists for taking time to be with us today and let's go on and kick off our interactive part of this super session. I'd like to start with what are the drivers behind wanting to look at intelligent networking are these more business or technical oriented and do you think it is the operational efficiency that we can achieve or is it making the network itself more intelligent. Lingli, could I ask you to kick off with your first answer to this first question. Sure. I think for our case the drivers are both and even more. So for China Mobile our business challenge the number one challenge is how to manage efficiently the largest mobile network in the world. We're serving almost one billion customers with five million base stations and for which operation and retainers we are hiring about 60,000 engineers and spending annually more than 20 billion US dollars in total. So to address this challenge our solution is to design and deploy a scale autonomous networks which drives the realization of end-to-end intelligent capabilities enabling the use of artificial intelligence within the closely controlled loops. Some of our use cases include intelligent 5G service assurance which reduces the network orders and increases the efficiency for abrasion and retainers and making our financial department very happy because we reduce the operation costs. And another one is optimized configuration management of 5G radio network and also automatic provisioning of enterprise network services which improves the coverage and making our customers both individual and enterprise customers satisfied. And in addition to that we also deploy AI driven energy efficiency which reduces the power consumption of our 5G base stations by almost 25% and that makes our network intelligence transformation really important because it actually contributing to save our planet. Right. So I'd like to add on to that so Verizon has a somewhat different challenge because not only do we have the wireless network but we also have the wire line network so we have a mix and a heterogeneous network that we're maintaining and of course we also have a global presence. So for us our challenges are certainly business challenges and operational challenges so we are looking for the intelligent networks to really address how the operations people are supported and get more efficiency out of the engineers that we use to maintain our networks but also of course use the networks more effectively and more efficiently. And I know Steve you probably have something to add to that. Yeah I would add on to that Beth I agree 100% with both of what you've said. There's also a complexity factor that we have to consider here as well. As our networks become more resilient and we have layered networks upon our networks it becomes more challenging to identify issues in the network and we find that AI ML helps us to narrow those points down without flooding our operations teams with too many messages that they can't use and focusing on the point where it needs to be solved and we see over time that this will have to become more and more integral to increase that overall customer experience by leveraging metrics more and more and applying it across different layers in the network itself. So what I am hearing is improved efficiency but also make internal operations folks happier improve customer experience and save the planet. That is what we're up to as we look at this transformation that's amazing related to going back to what we looked at in the presentation the EU AG just did this survey and what I'm interested in is not the sort of obvious results but what was most surprising to you that you found out through this survey activity. Steve could I ask you to start there? Sure, I'd be happy to Heather. What actually surprised me is if you look at the numbers and you look at the results one way you can say is there's not as much AI going on as you expected and especially in some of the areas like autonomous networks and things like that but I actually found it interesting to say 20% said they were fully autonomous or highly autonomous. That number actually surprised me because given the fact that it's challenging to get to the data it's challenging to find AI ML resources that can work on the data and it's challenging to process that the operations team through automations and systems like that can actually set up true autonomy that they were that far along. So it was pleasing to see that actually the 20% seems small but in my viewpoint that's actually further along than I expected because we hear the buzz about AI all the time and everybody's doing it but the reality is when you actually start looking into it and see who's doing this there's not as many people that are actually doing AI as we would expect and we hope to see that improve over time through a number of different projects and especially open source projects that can help to accelerate that. Yeah, Langley did you find anything surprising about the results of the survey and hearing the voice of your fellow operators? Yeah I would like to say what a coincidence What's the price as most is also what we found out with the road data and after we compare it with a more in-depth analysis is that we found out that the industry currently actually lacks a basic consensus on what is the intelligence transformation of the network or in other words the goal and the path of how to build an autonomous network despite the fact that many industry organizations and companies as I said have launched related work and also like operators China Mobile have also included autonomous network construction in our own five-year deployment plan but in fact through this survey we found that the interview is actually did not fully understand what the ultimate goal of network autonomy is and for example we found out that two interviewees said that their company's network autonomy level or intelligence level has already reached level 5 the ultimate goal for full autonomy but when we later combined their feedback with extra information about their application coverage or roll-out stages before we make the comprehensive judgments we found that at least one of them which represents 50% in total actually does not meet the standard for level 5 which states fulfilled or domain and universally capable of self evolution and we believe that this is a very good reflection which shows that there is actually no well-established industrial consensus of what it will look like or the ultimate goal for building network intelligence and in view of that we would like to address this issue because we have already included in our 5 year plan so to kick off the practical application in our production network we have to develop an implementation architecture and level based evaluation system which are specific to our own operation and retainer systems to guide the transformation of our network operation retainers transformation in all of our 31 subsidiaries in mainland China and we introduce autonomous capabilities at different levels of the network operation and management system and conduct evaluation one or two times annually to trigger a iterative evolution and the result for the very first duration which is partially down and the first half of this year is quite encouraging and at the same time we are fully aware that there is still a long way to go for reaching the level 4 go for our network and we firmly believe that only by joint efforts with the industry that we can actually continue this journey so we summarized and published our thinking practices as well as corporations adjustments and the white paper in July this year and further contributed the generalized methodology which we believe would be also applicable to other service providers and also our industrial partners for their references in the form of technical specifications to standards organizations including TM foreign, 3GPB I2T etc and that is what we have found out the most surprising to us and that is what we are react and tool it and trying to address it and I hope and it will also be of help to our yellow and service providers as well and very welcome other partners to join our practice in building advanced levels of autonomous networks together. So it sounds as though both you Lingley and Steve were both surprised by the combination of maybe self-reporting optimistic at a time when the consensus of achievement is still I guess itself under question Steve anything you have to add back into that or Beth do you want to jump in? I'm good on that one Heather I agree with what Lingley has said in that Yeah Alright then let me go to my next question which is we've talked about where people might be or might think they are in the journey and even what does this journey even look like getting started or moving along the path in the journey obviously there are going to be challenges so what do you see as the biggest obstacles which we need to overcome as an ecosystem? So I'd say there's two sets of challenges one is internal to specific companies and I'll talk a little bit about them in a minute but the other one is an industry-wide challenge as Lingley mentioned I don't think there's a consensus about what in fact intelligent networking even is and as we discovered from the survey many operators are focused on the operational aspects and less on the on the intelligent networking as in algorithms and automation so and those are really two different components you really need to address both internally many of our challenges have revolved around as a telecom rising in various forms has been around for over a hundred years and many of our operational systems are they're not a hundred years old but but they're not young not young and you can put all your lovely GUIs around the core data but at the end of the day it's very, very difficult to rebuild those data sets and so we have the sets of data that were built or designed 30, 40, 50 years ago and we're pulling the data out but it's not necessarily data that can be really applied in the way that you would expect for intelligent networking because things have changed over time so that's one challenge another challenge is of course Verizon is made up of a set of a number of companies that have been merged over the years and each one came with their own set of applications and data and OSS-BSS and OSS-BSS systems so that's a challenge in itself just crossing through those silos and I know Steve you've found some challenges as well related to algorithms and data sets Exactly, they're very diverse they're very different consistency would be very nice because most of your time is spent in data wrangling that you'll have I think we've found at least three to five different names for Jitter across the different data sets it would be really I need to pause you Jitter is named differently within a telecom networking's own data like data systems it's pretty sad isn't it Exactly because we have a number of different partners that we work with and everyone comes up with their own little twist so you have to take those twists and put them back into something meaningful so again you can layer that up and look through the network layers all the way from your physical layer all the way up through your virtual layers to your applications and that's the challenges is going across those because most of the time when somebody calls with a problem they're not saying I have a problem on my optical network they're saying my application is running slow and that's where you start at and then you start going through the layers to figure out what is really the problem Yeah so Lingli you were mentioning the size of China Mobile's network and your subsidiaries are you finding the same thing? Are there this many different words for Jitter? What challenges are you seeing out there in the world trying to do this? Right First of all I 100% agree with what Bas and Steve have just said data has always been a challenge for us and because data is actually to us is the fuel of artificial intelligence and we have tons of different data as you can imagine produced as a result of our OM system each day that without a unified data standards system established these data could hardly add value to the network intelligent platforms and another perspective regarding data's challenge is that we found there's also a lack of effective data governance solutions inside our network we have different management domains like technical domains and data does not flow freely among different domains among different administrative departments that for us is quite challenging because AI actually consumes data especially with the use case that Steve has just mentioned we need to leverage and to end data to diagnose what are actually the broken part of the end to end service link so data governance is also important for us to establish inside our system and aside from the data perspective we also find out environment is actually we believe the number one challenge for us is to do our intelligent R&D because our network is so complex it's highly dynamic well not flexible and the complex environment characteristics and its uncertainty actually provides extremely rich scenarios for us to do the network intelligent we see a potential like blue ocean but at the same time the complexity of the system or network and its danger also slows down the innovation in network intelligence innovation because we could not use the live production network environment to do the R&D applications because there has been great requirements for our services well the research and development AI technology actually require experimental feedback especially those involved in reinforcement learning and deep learning and things like that and also all the machine learning algorithms are actually iterate quickly if the actual data characteristics and changes in the production network which means that it has to be retrained and optimized in our cases on the one or two months basis so without environment for us to try new algorithms and its revision or models on the production network is the number one challenge for us to actually to do their algorithms and improve them and consistently and we believe that it is important for us to build an end to end experimental network environment and we are also committed to build that one at least inside China mobile's rim and open it up to our partners and even academia to try and help us to do more innovative experiments in that environment before we can actually deploy them in our production network after proper verification and I'd like to add so and I have experienced with other A.I. type activities in a previous life I did work on speech recognition which of course uses A.I. and there's a lot of bias that can get added into the algorithms unintentional bias and also a lot of black box thinking which is a big challenge because you really need to validate that those algorithms are actually working and not just magic pops out so that's an additional challenge that it sounds like you're addressing as well by having an experimental dataset to work with so that's a very key issue that I think the open source community can really address by creating that sort of standard dataset to work against that would be anonymized and not proprietary to each of the telecoms Yeah, I mean it's the complexity of the network and the complexity of the bias we don't even recognize inside our heads it sounds like our challenges and because I think both Lingley and Beth just sort of recognized how we start to deal with this complexity and how we deal with not being driven by our own bias because we're looking inward rather than outward brings me to open source and collaboration so we have some projects that we've gotten kicked off let's talk about how you as operators are contributing to those projects and also how you would like other operators and the entirety of our ecosystem to participate and what projects should we be considering also starting at the same time and how do we use open source to help us accelerate this and improve Steve I'll take this one, thanks Heather so if you look through the overall flow in the process of developing machine learning algorithms you have devices producing data in an intelligent in a non intelligent and an intelligent network you have data collection data preparation and then actually building the models and if you slice and dice those down and look at them really the first two components comprise about 80% of your work effort so it's the collection and the processing the last piece is really more minor is the modeling process we'd like to see that flipped around because the more we can reduce those first two pieces the faster we can get to the real piece that we want which is building models so we can have insights that will drive our business and make our customers experience better so our ask out to open source is to start looking at the collection phase of it to generate the log files today those are fine you know systems process those but can we make it easier to process those log files can we put those on a message bus at the same time so it's easy to pick up that data and start using it that's one of the main first components the second piece is data granularity data is data today from these devices anywhere to 15 increments to an hour or longer and the problem is in these large networks and complex networks a lot of things change in that time so a lot of things get missed in that signal if it's been aggregated down so we're asking go to lower levels of aggregation and if you can't even step down to raw where it makes sense so the actual symmetry coming off of the device so we can make it faster and quicker and richer and then finally the processing piece of it right we have this plethora of different devices that are producing data they all name things differently they put them in different formats they have different metrics in there can we take that and say let's put some standards around that and can we ask the open source community come up with common naming schemas come up with common practices for this data and publish what the data means I can't tell you the number of challenges I can't move through by getting a data set and then saying what does this metric mean doesn't having value how does it relate or we've got 12 different log files from this one device how do we put those back together into something meaningful if that's published that makes it much easier to get the data scientists going and not have to convert a data scientist into a network engineer as well because then they have to have that skill set to do that there's projects out there today like project and the data project which are great starting points but last also is to go out there and say on these other projects that are not related to make it easier for the teams by doing these things up front so we can expedite the process and make it simpler for everybody to get to quicker insights yeah what I really like too is how that sort of the challenges of getting data from devices that is a little it's too aggregated and not quite fast enough combined with what Lingley you were saying about getting the data into the deep learning systems to train them on something faster than a month basis right it's we need more data more quickly and more continuously and so Lingley hearing both you and Steven talk about data and its timing together do you have something to add and also in the context of how open source can help right as I said earlier we see that it is an urgent need to build an open innovative experiment to environment for intelligent networks R&D and providing the basic infrastructure testing environment data sets and also providing input for a generalized intelligent platform which actually provides the whole lifecycle management for all the network intelligent applications and it is also reflected in the EU AG survey that actually the more advanced stages of the interviewees enter into network intelligence transformation there are more of them actually adopting the unified intelligent platform approach and for us we are actually building the intelligent network platform based on ONEP and also Kubernetes and several other open source projects and we are also integrating with more advanced ML ops from other open source communities and as you could imagine we also have an open source infrastructure for the network 5G network based on open air interface and also some of our other open source toolings so open source are of much importance for us to build the environment for the testing for the innovation and we also believe that some of the testing and certification program or efforts inside AFN could also be helpful for us to identify or broaden our view in potential potential intelligent application providers so for us especially for OAN we have a very restricted suppliers you know chain and it is also challenging for us to introduce advanced AI technologies because those suppliers for our OSS systems are not that capable of doing artificial intelligence especially algorithms innovation as you can imagine so if we would like to also leverage the industry and identify more potential collaboration partners and open certification programs that would be also of wider help for us to get our go I believe I agree with that as well I know that we are working on creating the Anakid Assured program which of course is tied into infrastructure and not specific to AI but it will certainly help go a long way and I think I want to put out a call to action to the vendors to support this as well I know as what came out of the survey is the vendors are actually looking at AI as well but I think that they are not as far along as the telecoms if I remember from our analysis and of course another thing to remember is that as I said earlier there are two types of AI here that are by specific to the technology of the network which in production you want it to be in minutes you don't want it in two months and but also there is the applying to the operational aspects of it so I know that we are doing some work around natural language processing and other types of things that aren't specific to the network but are still moving along with the design of improving our efficiency in the operational aspects of running a very large network All right is there anything else that any of you would like to add to this conversation before we do our panel Lingli, yes Yes, just I think it's about also what Beth has just said so our current practice has been most focused on the operation layer but in the whole picture we envision that there will be four layers of network intelligence the network element layer the domain specific management layer the end-to-end operation layer and on the top of that is the business operation layer which actually enables our industry or customers and also expose the network management capabilities if they wish they could control and do self-service as well and we start at the operation layer which is the middle two layers but we expect to achieve the ultimate goal at least for level three or four we have to also look at what would be needed to be added to the bottom layer the network infrastructure as well as the top layer how we could enable and expose those capabilities to our customers and we believe that there needs to be a consistent systematic design to align different technical evolution standards between different layers because ultimately we would like to enable a vertical closed loop which actually triggers by the customer on the top and triggers in internal like closed loops at their underlying layers iteratively and that would be the ultimate goal and for that case we are actually also working with multiple standards development organizations that are typically working on different layers and there is currently none of them are actually working on all the four layers so we have to indeed leverage in what I think that could also only be done by CSPs working together to drive a multi-STO collaboration which is not the loosely collaborative model as it is now we really need a traditionally federated collaboration model between these different organizations and in addition to the document driven standards specification development I think open source as the format of learning code would be having the specific advantages in enabling the interoperabilities that would be quite essential to enable that vertical collaboration Alright Mr. Casey I think I'm going to leave you with the opportunity for the last word here Thank you I appreciate that Heather so I would say in tying into what Beth and Ling Lee have said we look at this as a network intelligence of kind of in two different pieces and this is kind of reflecting what Ling Lee said also earlier and that is that you have the intelligence of the network taking care of itself and making sure it's better for self-provisioning, self-help all those things that keep the network and make it go stronger and better for the customers but we also look at it from the application intelligence view as well and that's the downward view on this and that is what does the application need and how do we share that information back with the network so it's not just the network taking care of itself but it's taking care of the needs of the application which then takes care of the needs of the user and we need to look at both sides of the application because it's easy to start with saying hey make the network better but it's also make the network better for the applications and change that over time based upon what the needs are so if users showing up you change the network so that it supports their needs and the applications are running as they are not working as much the network can change and do other functions that it needs to perform and make this dynamic fluid type in the ML to build that intelligence for it and let me add and it should be transport agnostic i.e. be able to incorporate wireless wire line and other technologies seamlessly so that the customer doesn't necessarily know what network the packets are actually going over as long as it delivers the experience is it over glass is it still over twisted copper pair it shouldn't matter to the customer yeah and at the end of the day I mean I love what you were saying Steve that it's making the network more intelligent to better serve the needs of the humans using the applications and I think that that is a great place to end I want to thank each of you for joining us on this panel today I want to thank again China Mobile for the presentation that preceded this panel and I want to thank everyone out there in open networking land who has tuned in on our virtual one summit and chosen to take in all of this great information so thank you to everyone out there this concludes our super session on intelligent networking AI and ML for one summit 2021 wishing best of health and good living to everyone who's with us today