HockeyStick #10 - Generative AI in Fintech
Miko Pawlikowski ???
Follow for coding, bootstrapped startups & breakthroughs in tech. Founder, Engineer, Speaker.
Today, we're we're talking about Generative AI in Fintech.
My guests, Mark Brouwer and Chris Kardell, the co-authors of Build Financial Software with Generative AI, collectively spent over half a decade working on all kinds of financial software.
If you've ever wondered how we got from coin purses 3000 BC to modern, AI-driven fintech - we've got you covered.
Thanks for flying HockeyStick!
Podcast
Follow HockeyStick Show and find the episodes below:
Video
Audio
Summary
Join Miko Pawlikowski ??? in this episode of HockeyStick as he delves into the world of FinTech with industry veterans Mark Brouwer and Chris Cardell.
Drawing from their collective 50 years of experience, Mark and Chris share insights about their new book 'Build Financial Software with Generative AI.'
They discuss the evolution of FinTech, from ancient coin purses to modern AI-driven solutions, the challenges of writing a book together, the importance of adapting to new technologies, and the intricate culture within the FinTech sector.
Learn about the unique obstacles faced by financial technology, regulatory impacts, and the cautious embrace of generative AI. Perfect for anyone curious about the intersection of finance and technology.
Transcript
Miko Pawlikowski: [00:00:00] I'm Miko Pawlikowski and this is HockeyStick. Today, we talk about fintech, not just the stripes and squares of the world, but the broader term that encapsulates both modern financial technology and coin purses from 3000 BC. I'm joined by two guests who collectively spent half a century working in fintech, Mark Brouwer and Chris Cardell.
Miko Pawlikowski: They recently teamed up to write a new book called "Build Financial Software with Generative AI" and I'm going to ask them all about it. Welcome to this episode and thank you for flying hockeystick.
Miko Pawlikowski: I'm obviously going to ask you about that space shuttle in the background there. what's the story behind it?
Miko Pawlikowski: bought?
Mark Brouwer: this house, 25 years ago or so and, the owners of the house had a small son and, apparently he was into space and they put this up and I didn't have the heart to take it down. it's a beautiful, piece of wallpaper. There's also some, stars [00:01:00] on the ceiling that sparkle at night. I think their son was very much into space.?
Miko Pawlikowski: Initially, I thought it was one of the Zoom backgrounds, but turns out it's the real thing. All right, so I don't always get team ups for books, so I'm always interested in the backstory of how the book came to be, how we decided, okay, we're going to spend a year working together on something that's going to be relatively painful and that we're going to have to spend our weekends and evenings on. How was it that you decided to write a book together??
Chris Kardell: I guess, if you're going to spend as much time as we have working on the book, like you said, on weekends, getting together and in our spare time. for me, I wanted to write a book because I have a bunch of opinions on things and I needed an outlet for them.
Chris Kardell: I, I do like writing. And so I always wanted to, try my hand at writing a book. Writing fiction wasn't, uh, coming very [00:02:00] easily to me. So I thought that combining some of my tech knowledge with my love of writing would hopefully yield some interesting results. so far so good.
Chris Kardell: And I,brought Mark in because I needed his expertise in the subject area
Mark Brouwer: I never imagined my name would be on a book as an author and Chris invited me to join them and share some payments ACH knowledge. to help in writing the book, and so I, I happily joined in.
Miko Pawlikowski: So whose idea was it to write a book about AI in financial software? Is that also you, Chris? Or did Manning talk you into that because AI sells very well right now?
Chris Kardell: manning encourage us to drive In an ai pointed directionBecause FinTech and what we were covering, seemed a little bit niche, we've always tried to position the book and the technology [00:03:00] in the book as, something meant for a broad audience and not necessarily just something for somebody who's interested in payments or, finance in general.
Chris Kardell: we were open to any suggestions that Manning had for, Improving the audience or growing the potential audience for the book, and obviously, as AI is really hot topic, they encouraged the use of generative AI and exploring that throughout the book, and so that is the main impetus for including it.
Miko Pawlikowski: Got It. So they basically talked you into adding generative AI to it. Sounds like it. They
Chris Kardell: didn't have to twist our arm much, but, yeah, it was at their, behest that we did it. The challenge that we have now is that we're trying to cover,a full stack development book and also, We were a little bit hesitant to cover, generative [00:04:00] AI only partly because it's so new, it's moving so fast, we're new authors, we were not sure how long this process was going to take us and, how much we'd have to be continually updating the generative AI portion of the book. it does present challenges. But being able to dive into some of the different products and, get that exposure has been helpful as well. So it's been real beneficial.
Miko Pawlikowski: who should be reading that book? Who is it for? Who is it not for??
Chris Kardell: our goal, at least for me was to,bring 20 years of development experience and try to distill that down into a book for somebody. Either somebody who is new to programming or maybe looking to transition their role in within the finance industry.
Chris Kardell: in my 20 years, I've spent it trying to grow my [00:05:00] skills, as part of a team in whatever that team was lacking. So if that was testing or, Java, COBOL, shell scripting, whatever we were using that we were weakest in, I tried to learn as much as I could about that subject and that was beneficial to me.
Miko Pawlikowski: But it also left me with, a wide array of experience with, not a subject matter expert in any one thing, but I did learn a lot. And so I wanted to tie all that together. all the things that I've learned and give people perhaps like a road map that they could use with the book. we present, a smattering of many different topics, the hope is that readers will find something that interests them and then, go off and dive into that subject. Sure thing. So for a lot of people, I think the fintech term that was coined at some point, I suspect by some [00:06:00] consultancy company, means things like this new generation banking.
Miko Pawlikowski: there's a few logos that, that you see all over the place now when they raise another series D or whatever. but it's a much broader term than that, right? part of your book, that in the part that's available, I like how you cover the different hockeystick moments for fintech. from like coin purses to what we have today. Can you talk us a little bit, through that, and maybe give people a little bit of a taste of what they're going to see in the book, the different breakthroughs, important moments for fintech across the ages.
Chris Kardell: we wanted to present fintech as not just something, because fintech has been something that's been around forever. And so we tried to think back as to what might be some of the, earliest examples of FinTech that [00:07:00] aren't necessarily, tech-related or using software computer.
Chris Kardell: And yeah, so one of those examples that we brought up is, just example of maybe the first time somebody thought to wrap all their, coins or whatever their currency was at the time in a bag, right? that was probably, rather than carrying it, handfuls of coins or shells or, rocks, whatever they were using,wrapping that up in a piece of cloth would,be,
Mark Brouwer: Amazing, right?
Chris Kardell: Compared to, what we might consider today, with blockchain and bitcoins and, cryptocurrencies and then. just all the real time payments things as they are now. finance is an interesting industry, there's a lot that goes on, everybody wants to make money and obviously. The technology that's getting that money from person to person or letting you cash your paycheck or buy something is [00:08:00] pretty fascinating. It's been around for a long timeThat comes with a whole set of challenges to it.?
Miko Pawlikowski: so that's
Miko Pawlikowski: A lot of different, kind of elements there. But if we were to plot a timeline from, like the first, fintech thing to what you mentioned with peer to peer and crypto and whatnot.
Miko Pawlikowski: So what, we're starting at 3000 BC with a coin purse. What's the next big breakthrough that needed to happen? By the way, I'm happy they didn't like patent that in 3000 BC. That would have been a mess.?
Mark Brouwer: Yeah, good thing. There's lawyers Came after that, patent expiration would be out by now.?
Chris Kardell: So then what? I think you mentioned a cash register. Was that being the
Mark Brouwer: next thing? yeah. Perhaps the cash register. the FinTech covers, a broad range of things. So it's not just about money, it's also about the automation and efficiencies that we look to gain from the technology as well. we can look at the cash [00:09:00] register or, drop boxes where you might be able to leave your check overnight and have it collected in the morning, way back before we had ATM machines. Are different examples of the way that the industry just tried to address that, removing that human element where people make mistakes, so how can we help them do their job better? not necessarily replacing people, but just making them more efficient, which I guess also ties into you know what we're seeing with ai and some of the concerns right about it replacing us and doing our job and really all these things are just tools to help us work smarter, more efficientlysome of these things are driven by laws that The government and regulators are enforcing. We're always looking for ways to improve, ways to be more efficientIn the writing of checks, that was a big deal, many decades ago, just [00:10:00] being able to access your money to, make a payment to, another business or another person by, signing a check, handing it to someone or mailing it to them. that was a big deal. And, ACH, of course, that came around in the late 70s, where you can, use, electronics.
Mark Brouwer: You can use software to transfer money from one person to the other. And, of course, today, we have, many different,technologies that are vying for, top dog status, Venmo and Zelle and Cash App. there's a lot of different, things. The Fed is now involved in starting a same day ACH type of payment, where it will happen immediately or within a few hours rather than waiting a day or two, which is the standard ACH system.
Mark Brouwer: there's been a lot of changes, a lot of attempts to improve the system, make it more efficient, make it quicker.
Mark Brouwer: But these things take time, right? We're still using checks, although I certainly use, a lot fewer checks than I used to. I write [00:11:00] maybe one or two checks a year, as opposed to many years ago when almost everything I did required me to sit down, get out the checkbook, write a check, plop it in an envelope, put a stamp on the envelope, whereas now I can just go online, set up a recurring payment,to happen via ACH, and it happens.?
Miko Pawlikowski: Some people have said that ACH is going to go away with these new, payment apps, and ACH hasn't gone away. In fact, it had 80 trillion dollars in, volume last year, 10 years before that, it was half that it was a 40 trillion. So ACH is still growing, even though it's a slower process than some of these other technologies, but it's still quite viable and, still has a place now and in the future, but it's going to coexist with some of these other apps and I think some of these other apps and technologies, may go by the wayside, but that's to be determined who's going to be the leader.In your [00:12:00] book you mention banker's hours.which I found it was an interesting concept. how does that relate to this?
Chris Kardell: Yeah, I think, we mentioned bankers hours right as another example of the need for the industry to, move forward and, just grow and again efficiency and, Processing. So it used to be, when everything was needed to be hand done or processing wasn't there that you would have what we refer to as bankers hours, which from a consumer side, it was a derogatory term or where, you only had a certain amount of time to get to the bank every day.
Chris Kardell: So if you needed to, run your check or paycheck over to the bank to get it cashed, you had, a certain period of time, which was usually inconvenient if you're working a nine to five type job, or if you're not within those hours where a bank [00:13:00] might be open, you were just out of luck.
Chris Kardell: And Obviously, the way things are today, everything is,24 by 7 by 365, we're online all the time, you're able to go buy stuff anytime you want, There was a need for the industry to meet that availability requirements.
Chris Kardell: And so we talk in the book about the decline of bankers hours and, whether it's, just branches being open or allowing people to cash checks through an ATM machine or deposits through an ATM machine or now just Taking a picture of your check and having it appear in your bank account.
Chris Kardell: You don't even need to go to a branch for these types of things so we just talk about the decline of that notion of bankers hours. I don't know if my kids will ever hear the term, bankers hours, unless I bring it up to them now, because it's just not the same as, not that I'm that old, but, when I [00:14:00] was cashing my checks from my first job, I, it was something that we had to go into the branch to do and work on that.
Chris Kardell: So work that way.?
Miko Pawlikowski: Yeah. They won't probably recognize a check.?
Miko Pawlikowski: it's like that icon where you click and there's the floppy disk and it says save and people start asking, what's that? What's that supposed to be? Why is that?
Chris Kardell: I saw an article or a post about that floppy disk where someone thought it was like a vending machine with a drink in it or something like, and they didn't understand why, how that related to, saving something. So yeah, that's exactly right. we're long past the days where. Those discs are recognizable by kids, half the machines now don't even have, CD drives in them anymore.
Chris Kardell: All right.
Miko Pawlikowski: So let's switch gears a little bit and talk about
Miko Pawlikowski: Fintech in general.what makes it special? What makes it different from just general, [00:15:00] software development?you obviously collectively spent half a century doing that, so you should be able to answer that.?
Miko Pawlikowski: what makes it different??
Mark Brouwer: Yeah so the thing about Fintech is that it's very, very large .
Chris Kardell: in, except in opening an account, in, processing, thecalculating the interest on the account, assuming it collects interest, the processing for loans, there's just a huge amount of different laws applying to loans and, and a lot of that code just is enormous so it's not something that can easily be rewritten.
Chris Kardell: You can't take a bankingapplication and rewrite it in a few months or rewrite even most of it. And the same goes for the UI: there's just so many different parts and elements to the FinTech that you can't just rewrite in a quick amount of time. So [00:16:00] the issue with that is that some of the technology in FinTech is quite old and just gets added onto and added onto, over time.
Chris Kardell: And FinTech may be a little bit different than some other,technologies, where things can get reinvented and redone in a short amount of time. some of these companies: Fidelity, Vanguardbanking companies, they're all very large and there's a lot of barriers to entry So it's very difficult to start a new company and to get going. Now, lately there have been a number of companies getting started, Square and, other companies. They've done very well, but they're still, I don't want to say fledgling, but they're still getting started.
领英推荐
Chris Kardell: And, there's a lot of competition there.
Miko Pawlikowski: So is that really unique to the FinTech? Because when you were talking about it, I was thinking about like military bases still running on like the, Three and a half inch [00:17:00] floppy drives. Same with like infrastructure, the things that go with the motto, "if it ain't broke, don't fix it". And, then they age and age, but they keep working and, the cost to replace them is very high.
Miko Pawlikowski: is that the same or is it different for?
Miko Pawlikowski: Fintech
Chris Kardell: I think it's the same. you touched on a very important part of why it's so difficult to either modernize or update, FinTech software. and when Mark and I are probably talking about FinTech and, he mentioned Square and Venmo and those types of tools or apps,?
Chris Kardell: They're important and have had success, but they also still depend. They're still dependent and interface with this older legacy code, whether it's, through many layers of APIs, Mark and I, our experiences, working with,what we call a core banking system, which is. as [00:18:00] he alluded to,opening your accounts, just having your accounts sit there and accrue interest in all those business rules and things that go along with that. at my previous company, we were at like 30 million lines of code per platform.
Chris Kardell: And that was just talking about the backend code that we were processing on, not including all the UI or API code that, powered everything. And so to get that rewritten and work, a lot of that code had been running since the seventies. And so when you're talking about, does it work when code's been out there for 40, 50 years running, It's essentially pretty, solid, bug free if you haven't messed with it, it's not likely to break,to try to rewrite that, the cost is steep and just perhaps, the military or the aviation industry, these things [00:19:00] where, where it boils down to the cost of failure far outweighs, the cost of rewriting it, or, it gets tough, right. It's a tough decision for management. And I think that's where the industry is so challenging for probably that management layer where,you're talking about code that's still running out there from the seventies, how do you attract, new young developers into this industry where, they could go work on cloud or microservices or, all these other technologies, which are also within FinTech,but FinTech also has to balance all that cool new stuff with the stuff that's running on, mainframes from the seventies. it's an interesting balancing act, I think,
Miko Pawlikowski: Yeah, that makes me think of Windows XP that, if I remember correctly, was one of those like massive companies or maybe a government body that required Windows XP for [00:20:00] longer than they were planning on keeping it on live support. Microsoft did end up extending the support just because people were using it in the systems that have run for a very long time. But yeah, I imagine the issue of what you mentioned, eventually that code does need to be replaced. So this difficult balancing act is probably like a daily bread occurrence. What else makes Fintech unique from the technology point of view?
Chris Kardell: in my experience, it is that you have old code, that's running, like I said, from, something that was written in seven, in the seventies that you still may need to maintain or update because it has yet to be, rewritten.
Chris Kardell: So it wasn't uncommon for me to go into code, that was written, before I was born, and we're going to go into code and look at the, [00:21:00] the author date or some of the initial, changes in the comments and find, that yes, the code's been out here running before I was born,with code that's been updated and maintained for that long, a lot of the times it probably has lost the original intent of the author. So I think it can be eyeopening for developers, I think when we're all in school or if we've learned, taken a software development course, they talk about, comment your code so that, when you come back six months from now, you remember what you did,that takes on a whole new meaning when you look at code that's been out there for, way longer than six months, right? so the original intent is gone. It's been updated by so many people, good programmers, okay, programmers, maybe not so good programmers. A lot of the things that we look at and talk about today with, maybe unit testing and that type of thing, take a whole new meaning, [00:22:00] because you can see all these issues where maybe things that we learn about today or hear about today, have their roots, right?
Chris Kardell: you can see where people, instead of updating code, will spawn a whole new program just because we're so afraid to update existing code and break existing functionality rather than doing it, maybe what you would consider the right way, it's just creates a whole other. Program, which then has a whole bunch of things that come along with that. But, again, another challenge with that is the way that things are sometimes connected together. So you might have.COBOL code, or you might have RPG code on the back end, which then may run some type of, depending on the back end system you're on might be shell scripts or, job control language, or, all these [00:23:00] other, Pieces and parts are all in play.
Chris Kardell: So you have to have expertise or knowledge of all these different programming languages and concepts that may exist in other long lived software, butI feel like they're very prevalent in FinTech or in the finance industry, and it just makes for exciting, but also frustrating, times.
Mark Brouwer: There's also challenges in FinTech, with the fact that it's just very highly regulated. So in the United States, you have the Consumer Financial Protection Bureau, you have a number of other, US government agencies that, oversee what happens in the financial industry. 'cause. That's where the money is, right?
Mark Brouwer: So as a government is going to look to involve itself and regulate, many different aspects of, financial companies, that is something that continually has to be, addressed. So [00:24:00] every year the software has to be changed, to address these different regulatory requirements. and that's just in one country.
Mark Brouwer: So the United States has its own regulations. Europe has its regulations, and then all the other countries that you want to do business in, they're all going to have regulations, and they're probably going to be very different. So that makes for a great challenge in trying to create software and update software on a regular basis, so that it works in conjunction with the rules and regulations of the country that you're dealing in.
Mark Brouwer: Also, FinTech is a little bit different than, say, video game company, not to disparage video game companies for what they do. sometimes it feels more important for video game companies to exist than fintech. We all need to have a little fun, but, if your video game stops working, you're probably very upset and?
Mark Brouwer: your company may make the news because so many people are upset, [00:25:00] but if your bank is down for a day or two days, or you can't get access to your money, now you have a serious problem. Not only does it make the news, but financial transactions can't go through. And, to some tiny extent, the economy is impacted, obviously just, on a small temporary level, but
Mark Brouwer: when individuals or small businesses, economies are affected that way, that can have a major, detrimental effect to your company and obviously to the people involved. it's very serious. If you're going to make a change, that change better be right. It better not, negatively impact, the companies that you're selling this software to.
Miko Pawlikowski: How does all of this together affect the culture, it's gotta have an effect, right? All the regulations you mentioned, the higher stakes than video game company. code that was written before a lot of people were born. What does it do to the culture?
Chris Kardell: it's [00:26:00] challenging, but I always felt,at some points that that we were very close, a close knit group, with the team that I was working with. one of the things that we would have to do on a regular basis is, the part of the support either the support team where we're handling, the front line triaging of issues that are coming in, or, we are working on overnight support. And so a lot of the times. what we were involved in or what I was involved in would be if a lot of the processing and, a lot of the clearing of checks and processing of payments happens, overnight when nobody's, on the system and the system can work if something would happen, whether it be just bad data or code changes that were causing a problem, we would end up, resulting in, Calls to our tech [00:27:00] support division, if they're unable to fix it, then, we were getting calls that, one, two o'clock in the morning, whatever, waking you up, and then having to jump on and, code fixes, changes, troubleshooting. And sometimes again, depending on your area of expertise, if it's a loan or a deposit or, any of these types where there's a lot of business rules. you might be pulling in somebody else, so you might be waking somebody else up in the middle of the night.
Chris Kardell: SoI've always felt like the teams were pretty close knit. there has to be this sense of camaraderie or family where we're willing to, help out each other and work. So in that respect, I think it's a pretty cool culture to be a part of, but then you also, you do have part of culture as Mark has said previously, there's a lot of, big players in the industry, they're massive corporations. And [00:28:00] as a massive corporation, it's hard to be agile or maneuver change your direction, change. Some of the policies that you have in place. And that's probably true of any kind of big corporation or company that can be challenging for that. and so I don't think those go away.
Chris Kardell: But, the sense of teamwork that I've experienced over the years, I'm very happy to be a part of that.
Miko Pawlikowski: Do you share that experience, Mark?
Mark Brouwer: I do. the people that I work with are just tremendous and the amount of help that we all give each other and, It's just amazing. And we've all been doing this for quite a while. So it develops into a bit of a second family, because we all do, take care of each other. We know each other really well, and we interact really well.
Mark Brouwer: So I think that is, an asset to being in this industry, because a lot of us have been in it for a long [00:29:00] time. And even on other teams, I know the people on other teams have worked there for decades and they all know each other really well. They know their assets and the liability, so to speak.
Mark Brouwer: they know where they're strong and weak. So I think that is a good culture, when you're in FinTech is that you work together really well to, accomplish an end goal.
Miko Pawlikowski: Awesome. Your book is about generative AI, and in my head, it's a little bit of, contradiction with the fact that what you just described, with fintech, maybe being a little bit slower than everybody else to adopt new technology. What makes gen AI so special, that you see the need to write a book about itbasically two years after it became like a massive thing. Which is fairly quick.
Chris Kardell: generative AI is one of those things that I has, [00:30:00] skyrocketed in adoption by, the general public and users and companies as well, more so than I think there's more benefit to it thanwhat we've seen in the past in terms of other ways to,gain efficiencies in our work or within any industry. I think about things like, automated phone systems where, You try to say your account number or say what option you want and it never recognizes what you're saying and you're getting redirected. And all you want to do is speak to somebody and, it can be challenging. With generative AI. it seems that, have taken off so fast and certainly has challenges and along with that, but the benefits to it, Seem to far outweigh any kind of con that you could find, but in terms [00:31:00] of, FinTech adoption, or adoption in the industry, I think that's still, very challenging. I think that it won't necessarily be something that we can adopt until, confident that our proprietary information is not going to leave the building.
Chris Kardell: And a lot of what I've seen in, other, books or just talks, webcasts about generative AI, a lot of it, everybody's very wary about. Posting code and business logic into, something like ChatGPT and having it, sucked up for training purposes and potentially getting spit back out, but yet a lot of the use cases that people go over for generative AI are things like, code reviews or troubleshooting and, [00:32:00] gaining efficiencies in your code, which are all great, I think if, perhaps you're working in an open source type environment, you can use those types of things because your code has likely already been trained on the generative AI.
Chris Kardell: I think for fintech companies, they are going to be, Much warier about anything that we can post. And so our perspective in covering generative AI is to touch upon some of those, areas, but also to show the benefits of using generative AI, but perhaps not covering it and not using it in such a way as it currently is where we're just copying and pasting our code and asking, generative AI to give us an answer or help us with something. I found it a great tool to use when I need to enter code or have a [00:33:00] problem.I find myself having to try to rework an example of that code that I'm having a problem with,not just. straight copying and pasting it, but having to maybe rework it, take what I think is the problem out and, rework it so that I'm not going to paste something that is a, trade secret into our codes.
Chris Kardell: I'm not just worried about copying and pasting, maybe an API key or something like that. in FinTech there, we're very protective of the code, right? And just by virtue of trying to rework my code into something that I think I can post into generative AI, that I think is okay to post into generative AI,
Chris Kardell: oftentimes that just helps me get to the bottom of whatever problem I'm trying to work with. So I think the adoption will be there. I think companies are starting to do that, but [00:34:00] it'll be a long time before we start covering the use cases or using it as I've seen in many other. webcasts and trainings about
Chris Kardell: tools?
Mark Brouwer: And I'm old enough to think that two years is not a very long time at all. in the two or so years that we've, all become very much aware of AI, I think there's still a lot of work to be done on getting AI to work more efficiently, better. We've all heard about a number of stories about where AI is failing,
Mark Brouwer: now we have a new term that we're using, AI is hallucinating, sometimes in controversial ways, which you certainly want to avoid. and I think it's important to understand that we're going to have to take this a little bit more slowly that, we got to understand where this is going, what's allowed, what's not allowed, what's going to get you in trouble.
Mark Brouwer: again, I'll touch on FinTech and you don't want to get in trouble if you're a bank or a brokerage, you don't want to be [00:35:00] in the news because some flaw in AI that, has come your way.
Miko Pawlikowski: So the way things stand today, what would you say are the good use case and the bad use cases for Gen AI? Obviously that's going to be covered in the book, but given that, only four chapters are available, can you give us a sneak peek of what you're going to be recommending to use Gen AI for and what you'll be advising against??
Chris Kardell: from our perspective, we mentioned this early on, as generative AI being a better rubber duck is what we call it. So if people aren't familiar with, the concept of a rubber duck programming Basically, you have a rubber duck or any kind of thing that you want to talk to out loud, by virtue of talking to your problem out loud to something, even if it's not giving you any feedback, it just helps you clarify what's going on in your head and, [00:36:00] potentially get to a solution, which is probably more important these days where a lot of us are remote and it's not always easy to get in touch with, co workers like it might've been. I know that sometimes, my coworkers don't always use like a headset. I have to call them on the phone, if I want to talk to them, I can't just bring them up in a chat sometimes or have a conversation with them. So that concept of, being able to talk out loud and have a rubber duck is beneficial more I think now that we have this generative AI thing where I can enter questions or code, obviously code that I've, done something to, Keep it safe from posting, but I can have more of a conversation more meaningful feedback a lot of the time and I think that's very beneficial and so What we try to cover is not necessarily covering the use [00:37:00] cases from, a good prompt engineering standpoint, or, having, chain of thought or any of these other strategies, we're just trying to attack it from the angle of what we think a lot of FinTech people are going to do, we're busy, we're not going to want to spend a lot of time, crafting the perfect prompt or doing a lot of prep work. We want to just have a conversation,bounce some ideas off of it and see how it goes. there's a lot of great things to do with generative AI.
Chris Kardell: There's a lot of, Powerful things to do prompt engineering. There's a lot of different topics to cover with it. But,from, our experience in the industry is that a lot of people aren't going to be so interested in doing In the prepping of the model before and we're training of the model [00:38:00] before they ask it a question. We just want the answer right?
Chris Kardell: We want to get to from point A to B as quick as possible and so we try to show some of those use cases, some of the pitfalls of doing it that way you know some of the pitfalls of just using generative AI, without understanding some of that and then rewind and take a step back and go back into it.
Chris Kardell: from a different perspective. So I think the benefits for us or from our perspective that we're trying to present are more along the lines of this tool, being able to write great prompts is important, but that also leads you into just being able to communicate clearly with your co-workers,I think we have an example early on of saying to your co-worker: 'Hey, this is broke. I need it fixed. And then, as opposed to explaining what change did you just make? Why were you making that change? And, how is it broken? I think [00:39:00] I know that in the past I've certainly dealt with co-workers who would just come to me and say, 'Hey, it's broken, but I didn't do anything to it', only to come to find out that, yeah, they just checked in changes and then it broke, it was working fine until they made their changes, but they didn't suspect that anything that they possibly changed could have caused a problem. and so that's how, we try to feed that, type of conversation and tie that back into generative AI, using it as a tool, not to do your work for you or give you answers, because that's dangerous, but how can it, help you be more productive?
Miko Pawlikowski: The pitfalls you mentioned, I'm very interested in that. I love horror stories. Can you share maybe one about where things can go terribly wrong using generative AI??
Chris Kardell: I think that using [00:40:00] generative AI can go wrong now, especially if we're not careful. I know Mark has dealt with it in the fintech industry and certainly in other industries, but we are at a point where we constantly, have tools that provide,static code analysis for us to look for vulnerabilities, right?
Chris Kardell: And there's this idea of this OWASP top 10 vulnerabilities for any kind of web applications, but that's expanded to, not only web applications, not just, using Java or front end programs or front end languages. Part of that is also using COBOL or RPCA, they scan everything, right?
Chris Kardell: We're intimately familiar with, having to resolve potential cross site scripting or SQL injection type issues. But now, also, there are just a number ofgroup that comes out with those web vulnerabilities. now also has vulnerabilities for large language models. [00:41:00] And so some of the things we need to be careful of now are, possible, hallucinations from a large language model or generative AI, where it could be giving us, either a node package or NPM module that doesn't exist, but attackers or malicious users know it doesn't exist and we'll upload a, malware into NPM or whatever other package management system you have. And so if you're not careful and you just, copy and paste code, and into your thing, you could be into your program or IDE, you could end up downloading some of this malicious code and then getting it out into your application. And so those are,a new twist on, things that people should be aware of in the FinTech industry where malicious users will, enter, maybe a bank website, slightly misspelled or something and harvest credentials [00:42:00] from that, that's a kind of a constant thing that people are fighting against and, same thing with other,misspelled or misnamed packages, have always been a problem.?
Chris Kardell: But now, we have this other avenue for attack, where generative AI is giving us, code that we think, 'Hey, this would be great to test', or 'I could use this'. It doesn't really exist. People are taking advantage of that. So there's a whole slew of things that, we have to be careful about.
Mark Brouwer: Yeah, I think generative AI,is good for, obviously generating code, and also it can be used for doing code review. I would say that it should not be used for both on the same area of code. So you may want to save a lot of time, and possibly,prevent some bugs by having the, generative AI create the code for you.
Mark Brouwer: But I would say that you need to do a little bit extra code review [00:43:00] if you have it, do that. we've talked about how AI can hallucinate. I don't want it hallucinating when it's generating code. I don't want it making faulty assumptions that a human may never make, but AI might make.
Mark Brouwer: So you may save 20%, time and efforts inusing AI to generate code, but you're going to have to spend a little bit more, doing code review of that code. it's still a net benefit using generative AI, but I think we need to be really careful in, just relying on generative AI to either write the code or review it, we need to, also still involve the human and not rely too much on it. So there's a little bit of a danger in that. Hopefully we don't go down that road.
Miko Pawlikowski: So where do you see it going? Do you expect that FinTech will follow what the rest of the industry's doing and just adopt the hosted models, do you expect there to be [00:44:00] more of the local models that will become more I guess prevailing, because of the fact that you can host it on your hardware and you don't have to trust OpenAI to not do dodgy things with your inputs. What's your prediction? Where would you like it to go, and where do you think most likely it's going to go?
Mark Brouwer: I think FinTech is going to, use its own hosted, instance of generative AI, it's not going to, share it with anyone else. There's just too much proprietary information and too much danger in trying to share that with other companies. So FinTech is going to have its own copy of it, it's going to be tightened down. it's really amazing to know of all the different things that we have to do in the FinTech industry to try to protect ourselves and protect our code. banks and FinTech companies are constantly attacked by, foreign agents, and AI could just [00:45:00] open that up to even more problems, so we're going to be very careful with that.
Miko Pawlikowski: So what kind of generative AI are you using in the book then??
Chris Kardell: we tried to cover, ChatGPT and GitHub co pilot. both of those are, OpenAI, derivatives,we liked ChatGPT for the fact that it gave us something that we could potentially use,
Chris Kardell: No matter where you are or different restrictions that you might be under for, just installing things. So we thought that was a good option for, 'Hey, I need a tool my PC's locked down, I'm not free to install plugins. getting exposure to ChatGPT', and then, it being having the free version, available as well, it was a good option for users to dive in without, any kind of [00:46:00] overhead, we covered GitHub co pilot simply because that is another interesting tool that?
Chris Kardell: you can install in multiple, IDEs. So we use it in VS code, and then we also use it in JetBrains,IDEs as well. and obviously we cover that kind of use case, if your company is using generative AI, it does allow you to, have access to these, types of tools. that's a good one, right? They also claim that any code that it's analyzing is not getting uploaded or part of the model. So that was, the reasoning for choosing.two avenues. One where we have, deep integration with your IDE and another one where, you have that experience of being able to just chat and set up and work with, a generative AI model with ChatGPT,
Miko Pawlikowski: Are you planning on also adding some of the local models for comparison in the book, or is that going to be for the next book?
Chris Kardell: [00:47:00] if there is a next book, we might, cover some of that stuff.early on in the process, we were covering,some of the other ones like Amazon Code Whisperer and some of the other, plugins, but that again, just got too much to take on to try to show, a comparison of using GitHub co pilot versus what we got from ChatGPT versus what we got from, Bard or Amazon, we had different, examples in there.
Chris Kardell: And we tried to compare and contrast them, but since we're also covering, finance, ACH, Python, NextJS, was too much to cover, and without much benefit, so we may have to go back and, depending on how long it takes us to finish this book, there's a lot of talk about the large language models and
Chris Kardell: local solutions. But now we're also getting into like small language models. things that are more targeted towards, are your domain. So are there things that [00:48:00] we are definitely probably going to miss out on covering, but hopefully we cover enough to keep it interesting for people.?
Miko Pawlikowski: Before I let you off the hook, a question to both of you. Imagine that you can go back in time to the beginning of your fintech career. So respectively, 20 and 30 years ago, And, you can leave a note on your fridge to the past version of yourself. What advice would you give each other??
Chris Kardell: I'd advise myself to buy a bunch of Apple stock.
Miko Pawlikowski: That's a very good one.
Mark Brouwer: Your thoughts, Chris?
Chris Kardell: Yeah, that is good. I would also,advise myself of those, market things. but I got into FinTech by accident. I was out of,I graduated a long time ago, in 2002. I graduated college with a computer science degree [00:49:00] and I was. looking all over for work, it was not easy to think we just came off, the dot com bubble and, it was tough to find a job, especially where I was located. And so I ended up doing tech support type stuff at a bank. And my first real job out of college was, landed at a tech support role in a bank helping, write SQL queries for their software and doing a whole bunch of, internet support and all kinds of operational support stuff. And then, my first experience in FinTech and finance was get that finally land that job. And then six months later, the bank get bought out, which happens very frequently with banks. They're always getting bought and sold. And then everybody who was in that department was either let go or needed to integrate to another location. That's how I got my start. My note would probably just say to stick with it and persevere. It has [00:50:00] been a good ride for me so far, and I look forward to the next 20 years?
Miko Pawlikowski: that's a very good one. Mark, anything other than Apple shares, Google shares, maybe.
Mark Brouwer: Yeah, a wide variety of things, and importantly to hold on to them, don't tell them when they multiply by 10, because it's going to go up a lot higher. But seriously, yeah, I think Chris touched on, persevering. So a number of, points there.I don't know that this is advice I'd give myself because I also am pretty happy with the way things have turned out. But, I guess if I'm giving advice to others, then also I'd tell them to persevere and to just stick with it. Just believe in yourself, know that you have talent. There is turnover in not only FinTech, but any technology industry. if you get laid off, something happens to you, keep your head high and stick with it, it's almost never a reflection of you. It's just a reflection of some finances going on at the company at the time [00:51:00] and, dust yourself off and get back in there, find another job. That may be the one for you. You may prosper there. I've been fortunate enough to be able to stick with the same product for my entire career and that's worked out very well for me. I'm happy with what I do and with the people that I work with. So it's been a fun ride so far. Hope it continues a good while longer.?
Miko Pawlikowski: And that is the perfect way to end the episode. Mark Brouwer and Chris Cardell, everybody. Gentlemen, thank you very much.
Mark Brouwer: Thank you.
Chris Kardell: Thank you. Really appreciate the opportunity.
Breaking boundaries in FinTech's evolutionary journey. Insightful duo shares learnings.