good morning everyone and it's a pleasure to be here although slightly early for typically for for London citizens to travel out of Heathrow it's always typically that we're late so for this instance I'm early which is which is great and first and foremost you know the Fabian talk that we heard this morning it was it was great and I think it reaffirmed that artificial intelligence is in the here and now and Nvidia we're seeing that in abundance we've been on this journey from a development point of view for now five years if not slightly longer and part of the biggest fundamental change that customers are coming to us as in vidiians around the world and if anyone does know of Nvidia maybe you don't maybe you're thinking well I do know of Nvidia because I myself or my family or friends play video games and then they do you know great technology to accelerate those games that's absolutely true but I'm here to talk more about where we're focused on helping artificial intelligence and one
of the biggest things and biggest challenges is AI is fantastic of trying to make sense of the data because what we feel certainly in Nvidia and our customers who come to talk to us is that we have all this great data but we need to make ins we need to have insights into that to actually understand what that data can do for my business or how I can solve some of the biggest world's challenges but before I get into my talk I just wanted to show you maybe a quick introduction as to what Nvidia is doing in some of the use cases like we said earlier earlier on and it was mentioned is this stage is more you know has been brought to the table to say great that with AI is considered a future technology but we want to hear about the here and now so this video I just want to show you it's really just what Nvidia is doing and how it's working with partners and Rani's here from Airbus I'm glad
to say and this was of course it was planned there's a quick snippet of what Airbus are using our technology to do as well so can we play there the next slide please I am a protector exploring the expanses of space to keep our planets and all the species that call I'm a healer giving doctors the vision to create a in the guardian searching a mountain of data the sea of people to find the last jar powering autonomous machines that simplify caPSURE our greatest adventures and make us whole again I'm a visionary don't need steering I am the composer of the music for hearing and brilliant mind everywhere okay so so that was really just a you know 90 seconds show real just to show you some of the uses and use cases that's here now this is not a marketing video this is actually where NVIDIA has been working with probably some of you in the audience or some certainly companies around the world focused on AI and we're seeing this come through from
various industries now and just a couple that I've picked up on today and really what we're seeing a massive explosion of development in healthcare as you can probably imagine probably one of the areas that a I could fundamentally change a lot of our lives and family's lives and friends etc so for example a io state one of the biggest things that they see and it was a surprising statistic from my side was that strokes actually kill more people around the world than a number of diseases such as aids and tuberculosis combined so an actual shocking statistic and one of the key things that I'm sure all of us know is timing that early you can actually see and understand someone is having a stroke the more likely of a survival rate the problem with that is how do doctors diagnose the person is having a stroke very fast effectively and one of the uses uses that Ohio State has put into is using AI algorithm so they actually see in the u.s.
they see over 800,000 cases a year and the neurologists and if they're actually scanning the brain to understand if there is a particular on sort of a tumor or you know a disease in the brain how do they quickly diagnose that they've been using an AI algorithm and they've actually now in conjunction with a consultant managed to increase the efficiency by 81 percent so quick diagnosis which can lead to obviously a better survival increase in in survival rates and the list goes on so infrastructure for example and even into IOT so machine and robotics where diagnosing potential future false so GE for example where they're looking at machinery plants what's its statistical analysis of a machine failing rather than actually fail and have to send someone out to do a repair wouldn't it be better to do preventative analysis and again using sensor data and intelligent video analytics so where it's actually analyzing the actual machinery etc to say yes based on machine and deep learning they can identify that this this potential item or
Hardware industrial item is going to fail and send a an actual technician out to fix that problem so and Angie you've actually been using this and they've actually saved over six million dollars a year in just utilizing AI and put in particular deep learning so one of the key things as I mentioned at the beginning here is the onslaught of data because it's great if you have all this information so for example you know a radiologists taking an image of a brain etc or if it's of a machine etc but how do you make sense of that how do you actually identify exactly key areas and then with that amount of information how do you churn through that to then give you the answers and insights that you really need and you really need quickly we've seen since the 1980s and maybe some of you as well in the audience the the what was the rule called Moore's Law which effectively means doubling the amount of transistors every 18 to 24 months to help speed
up you know making insights and analyzing data that's really coming to an end and and he started in 1980s and NVIDIA brought our technologies out in 2010 around 2010 and since that point we've seen an explosion in terms of the performance what you can actually get from our technology that we work with and partner with for the likes of of IBM in terms of what performance you would expect to see so where you have this data and you want to get analysis and you want to analyze it as quickly and effectively as possible we're seeing over a thousand decks and we predict us over a thousand decks performance increase going out to 2020 so just two years away from what's available today so what you can actually get and make available in two years time will see a thousand decks increase giving you the performance effectively answers to your AI to your approaches to to diagnosis if you're running particular algorithms or running particular frameworks to give you answers effectively I think that the Nirvana
the the thing that we all want is real-time analysis where you can get answers to your questions now so you don't have to wake and run that algorithm or run that framework and wait a week in some cases actually it's a month where you actually want the results straight away so rise of GPU computing this slide is put it is essentially search showing that using typical traditional methods of answering questions and queries with with hardware architecture using the accelerated approach will give you that performance expectation also we're seeing the number of developers so if there's any startups in the room if there's any developers and data scientists and reach researchers we've seen a massive explosion in in people using this approach in using this technology to give you those answers and to be able to help you develop your particular solution so you can see there for example back in 2013 - today we've had over eight hundred and twenty thousand developers now utilizing this type of technology around the world and I do hope
this some in the audience here because I'll be certainly it'd be great to talk to you over a cup of coffee in the in the breakout sessions later on the other ones startups was a key sector favi mentioned a lot about startups and again in Berlin there there's a huge abundance of startups and organizations we've actually seen that as well so where companies have come forward so back in 2016 we had just over 300 today we over 2,800 startups working on accelerated computing and AI in particular using some of the frameworks in the middle and one thing I would like to highlight is if you are looking if you're investing if your organisation your business your enterprise organizations are investing in a I know there's no doubt that you'll probably be using one of the frameworks that's in that center column there so whether that's cafe or tensorflow will spark etc all of the good thing about what Nvidia do is we invest and we work alongside no the likes of Google the likes of
Microsoft etc to have all of those frameworks naturally accelerated using this technology using accelerated computing so by the time you download it and you run it as an open source stack within your environment it's already natively you can use the GPUs that Nvidia produce the other thing also is to get started sometimes you actually rather than making a significant investment what you want to do is just try this solution and we can this all this technology is available in the cloud so whether you go to IBM cloud or you go to any other instance where this AWS or its Amazonas or etc GPUs are available to actually in utilize and run these frameworks and try out your science so that's one other key thing wants to highlight the other one is inference so where you're on the edge so if for example you've trained your data so if you're already on the journey for AI and you're doing machine and deep learning so you've actually analyzed some of the data and then you deploy out
so for example with autonomous cars or some of the examples I've got here and unfortunately 20 minutes doesn't give me too long to talk about each one but I'll pick on one so KLM for example KLM over 215 million social media followers which is I think in any stretch is a lot whether that you know in facebook on Twitter etc etc so that's one way one approach to their customers can interact with KLM whether it's you know is my flight delayed or you know that I can't see these or any seats available except to rather than trying to phone and get through they're using social media to interact with KLM so to that they've actually hired over 235 social media employees that are they're looking and trying to respond to those customer queries because they want to offer great customer satisfaction the problem with that even though it's you know 235 people is great they can only actually identify an answer literally 1% of those queries utilizing chat BOTS developed using deep learning they've now
actually increased that from 1% to 100% so the deep learning frameworks are now looking going out and polling all of their medias and social news news groups etc and then coming back and answering general queries it could be a flight availability which you would know in real time or it could be more complex conversations such as I'm running late do you think I can get on the next fly and what it would do is it will analyze that that query itself not know how to answer that so actually point that to one of the 235 employees that they have increasing the customer service so again this is just one method and one approach that deep learning is being used in an in an organization such as Kalyan and there's we have over 60 65 over 66 customer use cases for different types of industries that's available today so please if you know if there's some here the actually would find more interesting to talk about come and see me I'd be happy to talk to
you about that so then leading on so this is my only I guess technical slide is well how do you do that you know how do you accelerate that how do you give answers to those insights in near real-time what are you actually doing so in NVIDIA we co-developed a technology called NVLink which effectively is a chip to chip interconnect so basically it allows you know organizations like IBM to put connect connections on their actual hardware architecture in their data center to speed up the throughput the capacity in that server or in that in if it's a multiple number of servers like I said in their data center so it can scale out so what this effectively means is putting the types of nvidia gpus in and then in our IBM server system you can then accelerate those frameworks that I mentioned earlier on with this cafe intensive flow and SPARC etcetera and it's using that art hardware architecture so that interconnect to talk from the GPU so the green the green blobs if you
like at the bottom the green boxes can then talk to each other as well as talking to the actual CPU within the server itself and and as I mentioned it beginning is I look after I manage the IBM business and work alongside IBM for Europe so you can see here there there power nine CPUs etc envy link is connected not only to each of the GPUs but also to the CPU and then it's also using the onboard memory so over a terrible are two terabytes of memory you can address so if you do have very very large data sets you know if you're if you're working on particular AI framework and you have a huge amount of data that you've trained one of the key things you want to do is run that through identity some image tagging may be actually do some natural language processing to actually take speech-to-text on in text to another language etc you won't be able to process that if very very fast this is one approach and this is
what we're seeing such a fundamental shift for companies using this type of architecture with the likes of IBM to speed and give delivered at their performance so in literally in a couple minutes I'm wrapping up now this is my last slide so how do you then take this forward what do you tend to you know how do you actually engage and where do you learn more because rise of AI is an absolute fantastic and this is my first time coming here and I'm blown away I think the attendance has been amazing I'm really looking forward to talking to some of you guys but from other conferences that we've spoken at and from customers and organizations coming to us is great we've learned we was just starting on this journey but we want to be able to then take it forward I need to hire some more data scientists or I need my data scientist to be trained and to be understanding this technology a lot better to leverage it in to actually make sense of
it more and videos been helping in this approach for now over ten years from a GPU acceleration point of view but much more recently with the explosion of AI we've now brought that conference to to Europe and I'm pleased to tell you that our GPU technology conference called GT C is actually going to be running in journey it's in Munich coming in in October and I certainly invite any of you that's interested where you actually maybe want to meet with peers and colleagues maybe you're a research or a developer and you want understand a little bit more about well what particular industries is this being used in how can I actually understand how can i leverage its technology how can I use it better this is actually this is the conference I would suggest you're more than welcome to to look at and take and take a you know a look through the website to take more insights and more information from it so it's got as I said they connect learn discover and innovate
you'll have all of the leading solutions that's going to be on display there be talks and from luminaries and maybe you can get fabian to come along and and give a talk alone as well and even maybe some of yourselves where you're working on particular AI framework you're working in a particular approach using deep learning we would love to have more understanding of what it is you're doing and maybe this is the conference you can do that so last year we actually had just over 3,000 people attend this year we're talking just over 4,000 so it's going to run from the 9th through to the 11th of October the website is there it would be great to see some of you come along and if you are interested in understanding a little bit more about that we have in the main exhibition hall where Nvidia is out there we've got one of our leaders for Central Europe that looks all of after our startup community Daniel Sarah Stowe is here he would love to talk
to you if you're a startup I mean even from a business enterprise point of view if there's any particular questions whether you've seen it on on the slides here or you just generally want to understand a bit more about NVIDIA and how it plays a role with AI I'm going to be on the booth and it would be my pleasure to talk to you so with that I would just say thank you very much for your time and I'll hopefully see you in the exhibition hall thank you