Everything as a Service: How AI Reshapes Financial Innovation

[Music] well good morning um my name is Frank I manage the cloud business from Huawei in the region um my topic today is how AI technology reshape Financial Innovation I want to start with a question a small survey how many of you have actually used Chad gbt well around maybe 30% um that's expecta and um now we are into a turning point uh AI being a buzz world for a while it's not new um especially recently uh the foundation model technology have really bring us to a stage that AI not only understanding the world not only pursuing the world but also creating something new not existing from the world um Huawei the the key technology Foundation model also called larger model uh was the key kstone Cornerstone of the whole breakthrough at Huawei we have uh investing in the larger model or Foundation model or quite a few years ago and and um we believe there any like just like any general purpose technology when human invented the fire invented electricity invented the Internet uh it's when you look back of our history those technology was not invented like one single day or one even one single year it's quite a been a journey that people adap to a new technology and convert that technology into productivity same into AI so today my topic want to focus in two things um first particularly when we adapt Foundation model technology into Finance what are the key critical factor and secondly I was more exciting to share some of the real use case the small progress when we adapted AI technology into Finance sector uh this are the most exciting part it's small Improvement but the whole journey of adapting general purpose technology happen with numerous small uh improvements uh in Huawei our strategy for AI in general is AI for industry uh I can recall Mr Gary when he talked about chity when we look at it's also a retail business or consumer business uh but in Huawei um the majority of our customer especially at the Huawei cloud is Enterprise and the government those those Enterprise business so with in those sector we buil a hierarchy architecture for our uh pangu larger model this hierarchy architecture can allow our customer and the partner to bring the own data set from their domain to train a spefic model that use the for their institutes and this applies to financi as well uh financial model is reimaging the digital Finance in uh three aspects first of all improving efficiency which means reducing cost um many new things Technologies happening for example in Bank sector today Mobile Banking can really provide a better a better experience comparing to a traditional branch and it work 24 hours a day and also AI technology can help new to generate new revenue and growth AI helps to uh find the complex pattern into amounts of unstructured data and also um model innovation in fact business model Innovation um we have we have working on a case with one of the largest bank in China we call the inventory financing you know when you try to credit a customer of a bank you use a lot of data on the book but not always the data tell you the truths so with AI you can actually find what goods in the inventory it's being sold it's being bought so you have a real real awareness of this business going on according to that you can build a model to create the customer um so at Huawei what are the key critical factor in terms of bring um Foundation model technology into Finance the first most critical thing is data high quality data in fact we always talk to our customer uh across all the domain that today to build a foundation model probably is one of the most uh sophisticated hardware and Software System in the world but what's the most valuable thing of the customer is their data because only the model trained or fine-tuned with the own domain data was not available for any public uh Foundation model will be crucial to train a model really can work and bring productivity to your workflow so high qu quality data will be the first Factor then the key three aspects we brought here as Foundation model first is innovative productivity following slides I want to share uh some of the user case how AI really increase the productivity and of course secondly is advanced intelligence how to really build a intelligent platform to really make your it more smarter into Finance um sector then of course abundant computing power is always required to train a model today require tremendous computing power and also I want to share some of the progress from Huawei Innovative productivity um so in terms of productivity we defined in three domain AI for data full link engineering the data set data engineering model engineering and application development how AI can help to improve the data quality itself then AI for business I just talk about how AI can help to build a new business model that you can never imagined before then AI for it here is a small demo that we work with one of the probably the largest bank in the world they have around uh 13,000 of data analysis but only 10% of those data analyst can write really a proper Circle the majority of those data analyst they came from a a finance background and then this became a bottom neck when um business leader want to see real time um the data inside the demo shows how AI can generate solve the problem when you talk with the system with a neural language but outcome is the graphic format and the uh and the Insight needed without a required the CLE skill sorry you don't need to understand any code to query the dat data you can just ask the platform using natural language such as how were the Performing loans doing last month the pangu financial model will first analyze the semantics and recognize that last month refers to June 1st to June 30th and the area is Shandon click customers with performing loans the data analyst assistant starts generating an SQL statement and shows the data source click ected private loans and the model generates an SQL statement and interprets the state keywords you can also have the model present the result in a graph that's just a real quick demo that how those small Improvement we using AI to our uh relitive work environment um also ints of productivity we using AI for for R&D assistant particularly in code generation here is a small demo that uh how to run a quick demo or it environment to develop something with AI today we the customer the partner we work with uh around many any area around 40% code was generated by AI um the second aspect is about advanced intelligence now um the really tricky thing intense convert AI into productivity is really how to convert the business expertise and the industry know how into your model and here is uh we talk about how to emerging intelligence or financial models is all the way from qna for for gener uh copy generation and uh uh your your business flow to create a financial model backed up with uh with NLB technology with comput compute vision technology with multimodality technology and uh um those those know how is exactly why we designed uh a hierarchy architecture because as a tech company like Huawei we can build a foundation model we can build a large Lang language model we can build a abundant power uh computing power but we are no expertise in finance we're no expertise in pharmaceutic we are now expertise in in uh in those verticals so what we do is that we we work hand in by hand with our customer with our partner to build this model to to using so the foundation model L we call LZ model was built by Huawei this model uh have the common sense trained by Abundant open data with open open source and also our our our data source from our particular library but the customer and partner from industry background help us to bring the own data set and help us to find tuning for specific tools um lastly the one of the last uh uh key factor is aboundant computing power while we just launched our Ascend AI cloud service Ascend is H own branded mpu uh a single Ascend AI cluster can provide 2,000 P flow from cloud and it's extremely stable uh we have tested with 30 a, cars of cluster uh 30 days continuous running the stabil stabilization rate is 90% of above uh this is all accessible from huawe cloud um lastly this is the high level strategic framework which we believe Finance model looks like first of all a scenario layer we really need to find out what other area are more suitable to start with AI start with now people always asking a question that will AI replace human well my answer will be it never going to happen but we as a human should be learning immediately how to harness AI technology and to find those scenario that human can interact with AI and pro bring productivity into AI then is the model layer the model layer is we uh the hierarchy layer we mention about the L l0 and the L1 industry SP model we have developed the pangu finance model and we welcome to bring any third party Finance model into our foundation lzo model and on the foundation layer of course the hog genous Computing framework which support GPU and our own mpu with a abundant computing power well lastly thank you I think the most amazing part is still remain unknown from the future and thank you for the opportunity thank