Greg Irwin 0:18

First off, nice to meet you all. Some of you I've had the pleasure in the past some of its first time. My name is Greg Irwin. I'm one of the partners at BWG. And basically, I moderate technology related forums been doing so for about 10 years. That's really what the community at BWG is. We partnered with NVIDIA on a number of forums and currently supporting them in terms of talking about AI trends. We actually scheduled this prior to the news here on ChatGPTs beta release. And the timing was, was opportune and really speaks to the work that the NVIDIA team does. So we've got Kevin Levitt and we have Tyler Condesa condenser joining from NVIDIA. But if you joined any of my sessions before, you know, I really liked spend a lot of time in these forums, what we think differentiate these forums is kind of the real stories about what's happening at different organizations. So this is going to be more about how financial service businesses and enterprises are pursuing and leveraging these tools, more so than even vendors in what the vendors are promoting. Not to say, I'm not going to get Kevin and team involved here. I certainly will. You know, they're here for awareness, they say enable some remarkable things more than one. So I'm going to be asking you to share a comment into the into the chat, and I'm going to start it right now. So we're going to do our intros. But as we're doing them, I want you to think about the one topic that you really want to hear from your peers on. Maybe it's specific to ChatGPT, maybe it's specific to LLM. Maybe it's just AI in general or certain use case for AI. But I'd like everyone to take take a moment, go into the chat window and share with us the one topic around a AI that you'd like to include as a point of discussion. As we do that. I'm gonna go here and let's start with some intro so you know who you're talking to. And I'm going to start here with Kevin Levitt. Kevin, do us a favor. Give a quick intro to your team over at NVIDIA. 

Kevin Levitt 2:38

Yeah, of course. Thanks, Greg. Hello, everybody. Welcome. And thank you for investing your time with us for an hour here this afternoon. I'm Kevin Levitt on the industry Business Development Lead for financial services at NVIDIA. So I help set strategy for NVIDIA in the context of financial services ecosystem, and then help bring that strategy to life through our ecosystem of partners focusing on education enablement, what are the workloads and use cases that leaders and financial services are prioritizing today, when it comes to accelerated computing, whether that's in the realm of capital markets, asset management and wealth management, fan banking, either on the commercial side or retail side, and certainly insurance as well as fintech. So definitely consider us a resource. We're agnostic as to where you consume our accelerated computing platform, which is full stack everything from the hardware to the software and application frameworks. To get to market quickly with AI enabled applications. We just want you to make the most of the opportunity that AI is presenting today. And we'll talk more about that in the context of our survey results and some of the benefits that our customers are seeing, and the industry is seeing it large. But thanks for hosting today, Greg, and welcome, everybody.

Greg Irwin 2:38

Hey, Kevin, do me a favor, you just reference the research you've done. Just give a quick overview of the work you've done. And I believe we're going to be circulating that after the event here.

Kevin Levitt 4:03

Yeah, sure. So this is the third annual State of AI and financial services reports that we've issued, we survey over 500 financial services professionals around the world to ask them about key trends or the application of AI in our industry. And you know, the some of the key highlights remain the growth rate in terms of the different trends for use cases. So to cite some of the data, as you might expect, the number one use case is the application NLP or natural language processing and large language models. Again, this was done before ChatGPT. So I can only imagine how many new use cases related to that technology has come up. The second most cited application is in the context of recommendation systems. So it's all about enabling the next best action so helping consumers make better decisions as they go about their financial journey, helping internal stakeholders make the right recommendations helping customer service agents and others as they go about their business. And the third highest rank application was actually an asset management around portfolio optimization. So we see a lot of use of our platform on the capital market side. And certainly not to be ignored. numbers four and five are both related to fraud detection, both for transactions and payments, as well as identity verification for KYC and AML. So that's just one of the highlights. I'm sure we'll talk more about it as the session goes on. As you said, this is less about NVIDIA and more about the opportunity for AI and our thought leaders are on the phone today. And we'll be sharing this report with everybody after the call. 

Greg Irwin 5:46

Kevin Jones sounds great. And now Kevin mentioned it. The reason we set up these, these webinars like we do, this is an open video, this is an open zoom conference call. So at any point, you can raise a hand, you can interrupt and say, Hey, I've got a question. You could drop it into chat. And the other thing that I really want to promote here is the idea of building a community. So we have about 50 people who registered and signed up and confirmed for this forum in data science and cognitive, cognitive tools at a range of FSI. I want you to think of this group as a networking opportunity. And the one goal above and beyond learning about each other's views and experience are stories around AI is to make one new contact across this group doesn't have to be Kevin or me. But one thing I would ask take a take a real make a real initiative that look across this Hollywood Squares that we've got here and make one new contact. And I promise this will be a worthwhile forum. All right, one of those Hollywood Squares is is Bruno Diniz. Bruno, do us a favor, give a quick intro to the group play. Sure,

Bruno Diniz 7:00

yeah. Yeah, glad to be here. I'm Bruno Diniz Ahmad, especially anything tacking based out in Brazil, asbestos, specifically in some Paulo. I'm an also an author here in the country with two bestsellers, published one on specific on FinTech and the other one and embedded embedded finance. I'm also a teacher at the academia here at the University of San Paolo. And I'm also a consultant on the FinTech field, basically advising international or governmental bodies, and also big companies on different different types of subjects related to financial innovation. So glad to be here.

Greg Irwin 7:39

Excellent. So Bruno, I will absolutely look to get you involved in I think that one of the questions, you know, please be prepared is what are some of the AI use cases that you're seeing across some of your clients? So please, please be ready. But before we go there, Efi, I'd like for you to please give an intro to the group.

Efi Pylarinou 8:00

Hello, everybody. I'm Efi Pylarinou. No, I'm based in Switzerland and I'm a global FinTech influencer. My background is Wall Street background career of over 20 years. So I've been looking at innovations in financial services globally, for the past eight years. I am a prolific content creator and I cover innovations in the business models in the strategies and how innovators are using these technologies in financial services. I'm delighted to share with you later on tonight what I'm seeing in the industry right now.

Greg Irwin 8:49

Kevin, do you want to get started with a story that that you've seen? And then I'll come to Bruno then I'll come to Efi and then I'm going to come across the group and ask people what, what's what specific use cases that you're seeing. I like to count, I like to cast them in the frame of a success or a challenge. So something where you look at it, where you had a goal, where you had a requirement, and by applying AI you were able to solve the problem. And by the way, thank you, everybody delivered a lot of good perspective, questions, ChatGPT a lot of customer service, a lot of confidence and trust around these AI models. So we're gonna hit on that, but maybe first we go to a success story. Kevin, can you share with us a success story?

Kevin Levitt 9:37

Certainly, actually, I would encourage everybody to register for our upcoming conference, which is our GTC GPU technology conference. It's going to be later in March because this story happens to be from a presentation that Capital One gave at our most recent conference back in the fall, and it ties directly into that top three use case of recommendation Mission Systems and how to use them effectively in the context of financial services. So Capital One faces a challenge and an opportunity like most of us in financial services and digital media, they're looking to optimize conversion and cross sales, which is how to make the most of your homepage banner placement. After one of your retail banking customers has logged in, either to your mobile website, to your mobile app, or your mobile website. And so like most financial services companies, after you log in, there's a premium homepage banner placement, and Capital One wanted to make that banner as relevant and as personalized as possible, as well as drive the most conversions and click throughs. So they leveraged deep learning and NVIDIA's application framework for recommendation systems called Merlin, to understand and identify the optimal placement to give any individual consumer after they logged in. The result was they were able to increase click to conversion rates by 60%. So six, zero. So if you think about increasing the optimization of your work, digital advertising in house by 60%, and what that can mean to your business, huge opportunity to deliver a personalized and customize recommendations. And it's not always about cross selling products, it could be presenting the right content to the consumer on their financial journey to increase customer satisfaction, personalization, retention, all of that is critically important. And I see a lot of questions about the application ChatGPT. Well, this is certainly one of them. When you think about generative AI, and the capabilities that NVIDIA and others are bringing to market is you can have the generation of personalized marketing copy, an email copy. And this is not just in the context of us as financial services leaders and trying to display the right content and creative to consumers. But it could be on behalf of your small business owners or commercial customers that want to understand how to optimize and personalize the experience and have the generative AI create relevant email subject lines, relevant email, copy, or even advertising copy. So hopefully that was a good example with some actual data as well as a tie into ChatGBT in generative AI for me.

Greg Irwin 12:31

I love it. Kevin perfect stuff. Let's go around and collect some more stories. But then I also want to encourage people, if I'm asking all the questions, it's gonna get a little boring. So raise a hand and an interrupt wherever however you you wish. I'll put some guardrails on it. So it's not a free for all. But feel free to you have you have an open open path here to ask questions and dig in. But as we continue to gain momentum, let's let's go to Bruno next. Bruno, tell us a little bit about how you've been leveraging AI for some of your clients, and maybe a use case and maybe we can talk about it a little bit.

Bruno Diniz 13:08

Definitely, yeah, there's one interesting case, I've been discussing with some clients, lots of possible possibilities with ChatGPT, specifically, but there was there's one company here in Brazil that have been able to leverage some of the capabilities of this to basically they are getting the engine of the they got the engine open and AI early like like, at the end of last year. And they've been feeding with some information because g4 education, which is the player here in Brazil, that they they've been feeding with their own content, because they basically produce content regarding and training regarding management, advanced management techniques, from everything from commercial to also HR, things like that. So they've been feeding this, the engine of open AI for like the previous three months. And they've been able to use first initially as a tool for their internal teams. And now I think that last the end of last month, they just released a version of it, we just called G for AI. So they're charging $10 per per user. And they have roughly now 2000 users springing you. But the thing is, is basically as it's more targeted type of of application, they've been able to so people that use this too, can both create some content based on on management and things like that, but also feedback to employees, reviewing business proposals, summarizing things that are already embedded in our themes, which is summarizing what will be the next Next steps for barriers, business meeting, the meetings, performance based evaluation based on reports on on employees. So all those things is interesting because it's pretty new, because they are leveraging the blue and he, and for some of the clients that I have, we are now talking about a next step of hyper personalization, and also contextual banking. And if you think about trying to use in conjunction with some suppliers of contextual banking tools, I don't know if you guys are familiar with players like fly bids from Canada, which basically tried to help on creating some static set of rules to contextually approach a client to offer some type of solution based on on on the context. So when you when you tie that to the capabilities of ChatGPT, you really can innovate in green things that goes first in the low hanging fruit here talking about advising, and even providing financial education in a basis that it reads really reads information on the client understands what's the problem that they have, and make an explanation specifically for a situation that perhaps the clients not even been able to assess themselves, and also to tax related questions. So there are some solutions in Brazil, we've been talking about, about that, which basically is a next step on some PSM twos, in all sorts of for

Greg Irwin 16:34

Bruno, I'm sorry, I'm gonna I'm gonna pause you there, because I think you're hitting on something that is a big hot button, which is, can you trust, the recommendations coming from these engines, particularly when we get two points of financial advice, tax advice, that is regulate highly regulated? And and, you know, and obviously, very sensitive? So what, what kind of guardrails Are you seeing being put in place?

Bruno Diniz 17:05

I was about to say about the challenges? And that's one of the challenges. So when you think about things, just like education in the case of G4, I think it's, you have less and less problems in terms of regulations around that. But when you're talking about financial advice, sure, you have some problems on that. So what we're discussing initially, with some institutions, there are already some types of tools. For example, a bank of Brazil, which is a big bank here in Brazil, they have AI to to help on basically, it reads, The punishing permission, the client, and the different types of, for example, expenses that they have. And they suggest a way for you to substitute, for example, some of those expenses and things like that, then also analyzing it, you can basically exchange different types of warrants for one law. And so things like that are really easy to assess when you go into something that would lead to liability, which, for example, financial advisory. It's it's the barrier that we have right now. And I think the industry is trying to solve that, especially here in Brazil.

Greg Irwin 18:20

Yeah, let's do this. I think I see a lot of questions here on ChatGPT. And we've got a pretty good sample set here. So let's do a live poll. All right. And the way we're going to do this is very unscientific, very non data science. We're just going to raise hands. There's no reaction at the bottom. And basically, all you do is a thumbs up. Everyone can see the thumbs up there. I'm going to I'm going to do it right there. So you can see it on mine. All you do is you mouse over the reaction, and you hit a thumbs up, and then it should disappear. I don't know at what point it is. Yeah, disappears after about 20 seconds. Okay. So now here's the question. For your, for your organization, or your specific clients, because I know we have some folks here who are in consulting world. What's the likelihood that over the next 12 months meaning 2023 You are going to pilot not deploy, but pilot uses for open AI or ChatGPT LLM for specific use cases for your organization? Kevin, there's one. Anyone else? Oh, Sanjay. Thank you very much, Bob. There we go. Damian. Yes, Martin. Yes. I gotta tell you a little bit fewer than I was thinking. That's interesting, Bob, Bob, I see yours. I'm looking here. That's maybe I'm not sure it's related for everybody because people are in different roles, but it's not as broad base as I would talk. Okay, well, let me try this. I'm gonna give it more time. Let's give it 24 months. What's the likelihood that over the next 24 months, not just this year, but next year, that you believe that LLN I'm driven AI is going to be in production in your institution 24 months doesn't have to be ChatGPT but at least something driven by a trained generative AI model. 123456. Okay, you know what, maybe this is a function of time, maybe this is a function of time, and just talk about how people are successfully establishing and setting up POC, because that's, that's the first step. So one, I'm going to ask for volunteer, if somebody can tell us how you successfully structured a POC, in your organization. Ah, I would love to hear it. Otherwise, you know, I'm good at asking questions. So I'm going to keep doing that. Anybody want to volunteer? Okay, I get to I get to pick then. And Efi, let it's time for us to get you involved here. Do us a favor, give a little intro talk about the use cases? And if he can put you on the spot and say, Well, how do large institution POC, a specific use case with with ChatGPT? 

Efi Pylarinou 21:14

Yeah, I mean, for for me, there's two main areas where there is not only potential, but they're starting to pilot. One area is capital markets. And the other area is what we call reg tech technology companies that are offering compliance as a service. So in the reg tech area, the company that I'm aware of that is already has integrated GPT 3.5. And they're piloting in two different ways. One way is, you know, a lot of regulations are not only complex, obviously, but they're conflicting. So a regulation in one jurisdiction, for example, or in a specific country might be conflicting with a global regulation, or regulation, that the investment world might be conflicting with data sharing GDPR and so on. And there's many examples. So they are direct tech company is using ChatGPT to add prevention? What if the bank fails to comply with a regulation that is related to the other one? And that's where they are training and getting meaningful answers in terms of the risk that the financial services organization is running, due to the complexity of these and in this is a highlight of how important it is it terms of what brands are asking these these models and now you're training them. So that that's one area and by the way, that company that I'm talking about is gone. To should go check them out. They're also doing specific pilots in in complex regulations around

Greg Irwin 23:47

Efi, I'm gonna ask you to drop that drop their link into the chat after you after you share the story. 

Efi Pylarinou 23:54

Will do will do will do I also want to bring an example that that is a FinTech that is older and has been bought by SMP global. And I see that SMP global could leverage them this company is called cancel. And I'll I'll drop the link to them because they're operating as a as a standalone company. And I remember them when they came out because we used to say, this is Desiree for financial markets because it had the plane prompt the like ChatGPT or Google and you could ask questions like give you an example. Which Apple suppliers share price goes up most when the company releases a new product. I mean, that's a very, you know, specific thing in financial markets question or, you know, you could ask an oil related question about the relationship of the oil and dollar. And it worked fantastically. And as you can imagine, the asset management world was was really interested in this. And there are they are using large language models, but training them in not only financial markets, but all the documentation, the huge documentation that exists out there, from earnings reports from all the filings, and so on. So they're, I, you know, I, I hear that they're starting to experiment and use these models, because, of course, if you use the ChatGPT, it will give you very general answers and not specific. So that's a very big area of leveraging the digitisation of the documents that is, that is that exists in gaining huge insights that are very important for the asset management business in the capital markets. Business. So though, those are the two areas where I'm seeing a lot of potential and interest is already starting, and I'm sure if our, if my colleague, Antonio Grasso was here, unfortunately, he is sick, he will be talking about what he's seeing in inshore tech, where they're using these open models in combination with telematics for for claim processing.

Bruno Diniz 27:07

I guess the interface thing makes a lot of difference, especially when you're trying to bring that to the to b2c, and even, in some cases for b2b, depending on who interacts with it. So if, for example, I paid, and also talking about the same line that Abby just mentioned, about stock markets and things like that, that before investor relations, for example, to be a, you know, very, very, very helpful on trying to respond all the different aspects of, for example, when the earnings call and there are lots of informations about the numbers, that the family on how you structure, your question your brand, you can really have very interesting insights out of it. So I think that's, that's basically, the game changer here is basically how the generative AI model is being applied as a new type of interface and the way they deliver the information, which seems much more intuitive than other other ways of saying, you know, AI, you guys know, have been extensively used in the industry. But but how it really changed these aspects. And and there's a quote, but also by, for example, they've the birch had been saying that, which is the big change in financial services not will not come when the on that when the financial services industry start using that, because they do that a long time, but actually when the clients the customers are using in adopting it, and I think that the mainstream effect that it brought after open AI ChatGPT I think now, customers are aware, aware of that. And there are multiple ways of trying to deliver information and make them create questions and have them responded in a way that is much easier than you know, from asking a cow and using a calculator in a bank, website or app trying to understand well, I'm going to be tired things like that, you just can you just can do that, you know, putting your inputs directly to the tool and having the response in a more structured way. So it's it's really definitely a game changer in this aspect.

Efi Pylarinou 29:30

Let me also add that you know, if you think predictive analytics, it if there's a prompt there, you're you're training your models, you get some quantitative results, right. And here, imagine that you have your team of whatever traders or analysts and each of them can ask so many combinations of questions and gain insights for that, it's just so much more powerful in a very simple and cost effective way. Whereas the ways that we have right now we really have to decide the one or two things that we're going to train for our predictive analytics, and we don't gain a lot of qualitative insights.

Greg Irwin 30:24

All right, I'm gonna wrap back here to Kevin, Kevin, Bruno, Efi, any closing comments you got for our group? And maybe just logistics to the follow up here?

Kevin Levitt 30:35

Yeah, start for the three of us just, you know, thank you, everybody, for the participation, the stories have been awesome. It's opened up a whole other line of thinking, you know, one of the things that we're paying attention to when you look at large language models and transformers isn't there, they're not tied to just language. We've actually done some very astute partnerships with folks like Conan Sears and others that are using transformers for synthetic data generation to help predict the CPI and the rate of inflation and among other outcomes that have great use cases, across all of our businesses. So just wanted to kind of drop that as another thought for everybody. It's not just purely about language and the creation of content, it could also be the creation of data to employ across a range of use cases. So we look forward to supporting everybody from accelerated computing platform perspective, as they go on this journey. As it relates to deep learning and financial services.

Greg Irwin 31:38

I think we can wrap here, Bruno or Efi, any closing comments, and then we'll send somebody off with send everyone off for some logistics and, and we'll get back to our day job.

Bruno Diniz 31:48

Yeah, I would just say about the challenges, you know, the base of the transformation here is fast. But the thing is, with, there's too many, some gray zones here in specially when I'm talking about a customer data and how to leverage these possibilities, you know, the customer facing aspects of it there, there are certainly some change challenges. But some of the examples were a saw that non sexy ones are very, are going to save a lot of money for companies as well and transfermate parts of it

Efi Pylarinou 32:22

to add is that one big survey in the US by CornerStone Advisors is showing that the big interest in banking is really conversational AI, rather than all the other items that we spoke about that are more around productivity and capital markets and asset management, we'll see safely, we can say that 2023 will be big piloting a year, at most.

Greg Irwin 32:55

Absolutely. Alright, so we're going to follow up with the survey results that Kevin was mentioning earlier, we're also going to follow up with everybody's names. And if you want to connect, come back to us here, we'll make an intro request. So I'm not just going to go and publish everybody's name. Of course LinkedIn is works pretty well too. So feel free of course, Tyler is on here at NVIDIA and Tom will be a point of contact if you do want to connect with NVIDIA. And I know I think Qatal are going to do some proactive outreach there. So if people do want have specific questions for NVIDIA, and a big thank you to the team at NVIDIA helping us put this together and bring and basically run this conversation. Thanks everybody. This is this is really cool.