How Technology Transformed Finance in a Turbulent Year

Jeff Macke:

Welcome to FinTech Corner. I’m Jeff Macke here with Brett Turner, Joseph Drambarean. We’re here to discuss all things AI, finance and the world right now. Brett, I want to start with you. We can’t ignore the fact that you’ve been a CEO. You’ve been a CFO, you’ve worked in finance for the last 30 odd years. Describe the environment to me right now because it always seems unprecedented, but this seems unprecedented in the weirdest way possible. We’ve had a run on the banks and then AI was going to save the run on the banks, and since then, interest rates have just gone completely bananas. It’s been quite a year.

Brett Turner:

It’s been a crazy year, and here we are at AFP, the Treasury Conference, the annual treasury conference, and a lot of legacy tech sort of surrounds us, but we’re doing something at a time where unprecedented in terms of this time of innovation, so not only all the circumstances crazy, but it just really highlights the fact that if you don’t have at the heart of innovation, if you don’t have great tech to try to wade through all that’s going on, it’s really hard and you’re caught flat footed constantly.

Jeff Macke:

Now you deal with customers, but you’re also running the business. How would you describe the environment out there? Because in times of uncertainty, you get clients who get kind of the Godzilla arms, they can’t quite reach their pockets, they can’t afford not to at this time. How are you adjusting for that uncertainty in this environment?

Brett Turner:

Well, yeah, I think the biggest thing is all that just needs to happen from a communication standpoint and just that you need just information and more data. So that’s what the more data and more information that you can really communicate and leverage off of is going to help you through all these kinds of times. And I think that’s, you see it and with those that just don’t have that, it’s really hard to navigate. So responding to, we’ve never seen a collapse, like what happened with SVB just unprecedented and then three other banks that really followed short.

Jeff Macke:

And when we talked in March, it seemed like that was going to be the trend. That was a risk that we didn’t foresee, and all of a sudden that kind of was ring-fenced as well as it could have been, I guess under the circumstances, but it hasn’t played out necessarily in a clean way. Now we’re at the highest rates that we’ve seen by some measures since 2007, which was a much different economic time. That was the pre-great financial crisis, kind of the waning days. It was a much robust economic time. Right now we’re waiting for a recession for the last three years that’s never come. Are you seeing spending? How do you talk to clients and how do you go out there as a businessman yourself and make plans in this environment?

Brett Turner:

Well, I think there’s so much nervousness, anxiety, the buzzword of the decade that’s happening, but that’s permeating and not just customers or even within the finance ranks of just lack of visibility, but even within your own employee base because everybody sees that you’ve got layoffs that are happening a lot. So there’s just a lot just constantly concerned, and then you’ve got world events is starting to take place. I think we’re seeing just a myriad of things in ways that are causing all kinds of uncertainty or volatility. And then how do you respond to that?

Jeff Macke:

Now, Joseph, how you’ve responded to it in one way. You’ve been taking basically the ChatGPT, the AI tools that have come out there, and obviously a chatbot is not a new invention, but the way it’s been incorporated over the last year has been exciting. It’s been incredibly innovative. How are you developing product and what are you looking for as you create these things for clients to really get them to be implementing this new kind of tech without being terrified of the security issues?

Joseph Drambarean:

Oh, it’s been a fascinating journey. So we tried to launch the first version of our generative AI solution earlier this year. So it was around the summer timeframe that we started to put it into production for all customers. And our perspective at that point was to focus fully on privacy and security because the feedback that we had heard not just from our customers, but the wider industry when it came to playing with this technology was well, how can we trust that all of my data isn’t going to be used to train the entire system, and how can I also trust that the answers that I’m going to get from this system are viable and that I can actually use in day-to-day business? And so that’s where we put all of our focus into, and I think that we did knock it out of the park in the first iteration in that regard because we created the first solution that was privacy focused.

We made it such that we did interact with some of the best tools like ChatGPT, others as well. Bedrock at AWS is another technology that has come out since, but what we did was we did it in a way where your data never leaves the Trovata environment. And additionally, we wanted to make sure that the answers that you get from this system go through the processing of Trovata so that you don’t have to worry about the math being wrong, for example, right? The whole issue that we’ve seen with these chatbots of basically the chatbot losing its mind, if you will, the inability to do simple math, two plus two equals four.

Now, what’s happened since then has been fascinating because that was our first iteration and what we’ve seen is an appetite from customers to use this technology almost like an assistant, right? Where better searching, better find findability, looking for needles in a haystack, random questions that you might have to spend hours figuring out the answer. You just on a whim try it in the system and it’ll find the answer for you. What we’ve seen is that over time, especially if you get past that first barrier, the Oh no, what is it going to do? Can I trust it? Once you take those steps and you start to see what it can say back, you start to trust it a little bit more and take bigger risks if you will, in the sense of from a financial professional, all of this is a risk, right? Adopting this technology is a scary venture. So as we’ve seen them take more and more steps towards the technology, the questions have become richer, and I think that’s where we have seen the opportunity for innovation because we can take advantage of the natural capabilities of our platform, whether it’s in forecasting or machine learning for predictive things. So where will this data go and marrying it with this generative technology where it can answer any question that you can think of.

Jeff Macke:

And how do you prevent customers? If you’re in a environment where one scenario you would ask it is show me risks, let’s run a bunch of hypotheticals. How do you prevent that becoming a weaponized version of this product that’s used against you? The best way to reduce your outflows would be to fire all the employees. That would be a solution that your chatbot might think was a great idea that probably isn’t that well thought out. How do you get in front of that in terms of getting people to ask it the right questions to create the right solutions rather than in fact just enhancing risk rather than ending up with a lot of good financial products starts as a great idea and ends up almost blowing up the system?

Joseph Drambarean:

That’s a great question, and it’s actually in line with a philosophical stance that we’ve had from the beginning at Trovata with regards to the use of machine learning. We have always felt that it’s important to always consider the true artist, the human, right? The person that’s making the judgment calls about the outcomes of the business and where the business needs to be steered has to be based on human intelligence. The benefits of AI are in the fact that it can do the mass computing, the mass scenario analysis and create a bunch of perspectives for you. But you can’t take the operator out of the equation. It’s similar to forecasting, right? If you’re looking at, Hey, what is an approach that we might take to protect our downside risk, but you don’t want an AI solution to just say, well, guess what? You could be saving 20% off your bottom line if you just got rid of all of your staff. What it’s not going to do is consider all of the outcomes. So I think that’s one of the things that you still need the human operator to be part and parcel in that whole equation, but when they work together, I think that’s where we’re excited because you get the creative capacity of a human with the infinite compute capacity of a machine, and when you bring them together, it’s Ironman, right? It’s Jarvis and Tony Stark.

Jeff Macke:

Absolutely. Brett, question for you. Everyone’s playing futurists with this. Jamie Dimon says that it’s going to create a three and a half day work week with this AI because we’re going to be so much more efficient. That’s been the promise of new tech for the last a hundred years, and it’s never quite panned out that way. The dishwasher did not in fact make homemaking obsolete. The future is never quite how it turns out. How are we going to fill our time? How are we going to fill our resources and really expand as we get better tools, as it doesn’t take us three weeks to do forecasting, but rather it maybe takes us three days if we’re going to do something robust? What do you see as being the next step of this, the next iteration of how we use this power?

Brett Turner:

Well, I think just even over the last decade when you look at the cloud and all this tech that’s permeated every aspect of society, it’s just created a massive amount of additional productivity and tons of leverage of pretty much every department that’s touching it except for finance, except for treasury, and except for accounting, they’re all three very siloed. So you got all these dynamics happening where they haven’t really had folks in these areas in the broader finance function, haven’t had their leverage moment. So in cloud, there’s a tremendous amount of leverage that you get in all of the older IT tasks of now instead of being on a defensive posture, it’s uptime risk management. From an IT perspective, all of this got weaponized and then you had this whole digital transformation. It caused a massive amount of innovation and productivity across the board.

So if you look at on the finance side, it’s sort of whenever you want to analyze something. So all the things we’re talking about when there’s this disruption of banks or world events or inflation, how do I take advantage of a yield environment, excess cash when the last 10 years rates money has been mostly free and there hasn’t been any advantage to get any yield off excess cash, so you just leave it in the bank. So you got to be able to assess some of these things or bigger things where we went through covid and how is that going to impact your business? Well, massive implications. So these are all requiring tons of modeling, tons of projections, and you look at the need for forecasting or you got to be able to start, nobody has a crystal ball, but you have to be smart and proactive at trying to look out there. So that’s a big part of managing risk, making decisions now based on what you think is going to happen. So I think all of that, being able to, whereas finance is so far behind the curve right now, being able to have this to allow it to catch up and have leverage and then start to go on the offense just like everybody else has had the advantage of, it’s going to be a game changer.

Jeff Macke:

It’s really two huge waves just crashing against each other where you’re making the finance guys, all of a sudden they’re not just an afterthought. For the big thinker that’s working, Jarvis plays the ego to Tony Starks it all the time, keeping them in checked and reminding him, oh, it’s rational and things. Now all of a sudden you’re super powering people that haven’t necessarily had that much control before. How do you see good founders, good companies really again implementing this so they don’t get too far out over their skis trying to weaponize this area of their company that before eight months ago, they never occurred to them to even really pay that much attention. Do we have enough cash to pay our bills? That’s just fine unless the bank closes. Oops. A little bit of a hiccup. I mean, I guess we’re asking kind of the same question again. What direction do you see this taking place that can control risk while at the same time weaponizing that while giving more power to the finance side?

Brett Turner:

I think two ways. I think you’re going to see, so here we are at AFP a treasury conference and you look at a lot of these older legacy solutions and they’re built on 25 year old, 35 year old tech. There’s a lot of on-prem solutions, there’s a lot of just stuff that’s not natively built in the cloud to take advantage of all this great automation. So I think there’s two things. You’re going to see massive changes in the way workflows are managed, particularly around how people do their jobs,

And you’re going to see folks who have the knowledge and who have the experience. And like Joseph said that the value of that input and guiding the tech is so important. So those folks who can kind of see the bigger picture a bit and how to really drive that to their advantage are going to thrive. And other folks who are the ones that are maybe doing more of the manual tasks probably, it’s not like they’re going to go away. They’ll have to kind of get retrained. They’ll have to focus on being able to get there and adjust and be able to leverage the technology as opposed to doing the things that technology is going to displace. So I think that’s a big thing. So you’re going to see, I think treasury start to really change, get more automated and then get subsumed into the broader corporate, the corporate finance function. And then you also look at some of the things that AI and automation are going to drive is also things that make it, especially AI and generative AI is going to, you take that and you start to get more of the and sign off in some ways, like validating that it’s accurate, but then in the hands of even non-finance people

And now this whole interaction of a CEO having to ask the CFO, who then trickles down to the team and somebody spending two weeks on some or a team of analysts doing something and then has to work back up. Now you’re going to see just kind of in some ways that cutting out the middleman concept, you’re going to see business folks do finance things with incredible leverage within a certain parameters, but they’re going to be able to do things and just get that information so quickly and maybe it’s just validating or maybe it’s so accurate within certain realms that they’ll be able to act on it right away. So those are the things that you’re going to see that’s going to be game changing in terms of the whole function.

Jeff Macke:

And Joseph as a product developer separate some of the hype from the reality because you’ve been able to get to market fast, you’re in there creating it. Talk to me about what the AI is actually able to do beyond just being a chat function that created all the buzz and happy birthday ChatGPT for a year ago. It was nothing. And now OpenAI is going to be selling employee stock for valuation and 90 some billion. How are you kind of validating, if you will, that the potential that is expressed with that 90 billion and driving it down to the companies as you develop product?

Joseph Drambarean:

That’s a great question because at the end of the day, we’re kind of in the maelstrom right now of marketing, if you will, and there are so many companies that are building on top of core technologies like ChatGPT or Google’s Bard or AWS is offering through Bedrock. They’re all popping up and obviously from an investment perspective, it’s one of the hottest areas to be investing into, but when you peel back the layers and you start to actually look at how customers are using this technology, you’ll start to notice that it’s actually an evolution. It starts with the smallest baby steps of trust, and that’s what we’re starting to see with real customers is that you have to participate in their journey. You can’t just assume that customers will come straight to this technology and trust it with their life. And so as a product development team, we at Truvada, we really focus on understanding what are customers trying to do with this thing, right? If we’re at the stage at this moment of technology where maybe they’re only focusing on this being a findability tool and it’s an extension of search ultimately, and that’s because that’s where they’re at, they’re only willing to cross that far.

Jeff Macke:

And just the popular press is doing it, no favors it, and just in what you also see at home, that was really kind of what worked with the ChatGPT, all of a sudden your homework became much easier. All of a sudden the idea of a term paper became kind of obsolete because you can just have it written for, you can do 500 words on Napoleon and I’ll edit it a little bit, and all of a sudden your 11th grade history paper’s done, these are the first functions of it, the deepfake and things, that’s what happens. It’s also the beginning of war insidious possible churn. Are people becoming paranoid of it or do you think that it’s becoming much more rationalist response because the early adopters often are the most reckless, the people who trusted this right away, maybe it works out for ’em, maybe not, but you guys are kind of tapping the brakes and the gas at the same time. How do you strike that balance with customers?

Joseph Drambarean:

And wait, the beauty of it is we are getting to see all of this in real time. That’s the beauty of this technology is because it’s interactive, because it is chat driven, you get to see the real interactions that people are taking with technology, and that’s a little bit different from you clicking around on a website and us kind of surveilling what you’re doing. This is real interactions and we get to see the cognitive approach that you’re taking, any problem that you’re trying to solve. Some folks that, like you said, are the early adopters and are willing to take risks, they might jump all the way to the future, which is, Hey, here’s a problem that I have. Just do it. Create a report, move this money, and we have to actually restrict them. We have to protect them from that outcome because there are all kinds of risks that they may be embarking into if they jump too quickly into that. And so that’s where we put our guardrails as a product team because we have to make sure that they move through this journey in a way where the expectation of what the outcomes will be is predictable. You don’t have to gamble every time you throw a question at this system and wonder, well, I wonder what it’s going to do this time when I ask for a forecast or when I ask for any particular thing. You don’t want to be taking risks with moving funds, if you will.

Brett Turner:

You can’t take risks when you’re trying to reduce risk. But I think that if you look back now, we take so many things for granted that when you look at the early adopter phase of tech, it’s been happening for a decade. So a good example would be the cloud. Now everything is happening in the cloud and we take

Jeff Macke:

It for granted. We used to just have some clauses and every company had its own just hardware. Now it’s in the cloud and we really, it’s not even something you think about.

Brett Turner:

People would say, oh, it’s in the cloud, and those discussions, nobody talks about it anymore. It’s just part of an extension of your everyday life. And then it’s also a huge part of the business construct and how businesses are run. Things are operating in the cloud, and I think in the same way, but when you look at cloud adoption, that whole journey to the cloud, it started with moving on the business side more enterprise type workloads because AWS peeled out from Amazon in 2006 and then that started, it was a lot of research projects or maybe development groups within big companies on certain projects, but it was sort of fairly nascent. I mean, it started to kind of grow under the, but nobody was using it for straight up IT dependency kinds of workloads. So then all of a sudden it happened when all of a sudden people were like, Hey, we could outsource our email server instead of being in our own data center, our data closet, we can move our email into AWS and we can provision that and not have to pay some outsource provider even. We can manage that very simply. And these services started to pop up enabling that in more protective ways. And so that was the tipping point of really enterprise in AWS, which really kicked a lot of this off, didn’t happen until May of 2013. So you’re tagging seven years after now. I think everything you see speeding up now

Because what the adoption curves and past or compressed now because everything speed, but I think we’ll continue. It’ll start with a use case. I think it’ll probably going to start with more something around forecasting and there’ll be something that, because forecasting is just such a religious thing in terms of how people do it, instead of arguing about how to do it or the model construct or certain things or even the assumptions or various things.

Jeff Macke:

With Amazon, you’re trading earnings statements. We’re going to hear from Amazon, Microsoft, Google, they’re all going to offer forecast as someone who comes at it from the other side, a more consumer, more equity side of things, the results are probably going to be good, their forecast is going to be conservative, and then the real forecast is going to be something they don’t really share with anybody. It’s how do we view that as if you’re a consumer of the news, if you’re not an AFP attendee who’s really wired into this, but rather someone who’s just trying to make heads or tails of it in a world where everyone’s talking about AI, how do you separate what you guys are doing, which is actually creating real functions, real application for it from just the chatter, just the blah blah.

Brett Turner:

Yeah. What we’ve done is we’ve really built the first true big data platform that’s leveraging all your bank data, and then we’re traversing that in terms of other finance data as well. So tru a trove of data. We’re building the smartest trove of data, the smarter that is, the more right now teams are being able to work with their data in a very flexible and lot agility, great speed, all of those kinds of things that they just can’t use with legacy tech. So all the underpinnings are positioning for them to do that fast and successfully in that way, and so they’re getting great gains from that standpoint. But then you look at that same underlying tech leveraging ai, it takes it to a whole different level because if there’s only so many buttons, two hands can push, right? There’s only so many hands you have. AI can just give you this certain extension, use the same tools that you’re getting comfortable with already, and then it’s able to really take the automation side of things to an entirely different level.

Jeff Macke:

Joseph, where is the most value being added right now? Everyone sees the pop culture things. We’re talking about term papers, we’re talking about DeepFakes. Tom Hanks is out warning his fans that he’s not in fact selling cheap dental products because there’s a deep fake out there, these nefarious uses. Where are you seeing the real development take place and what’s kind of something that people aren’t talking about?

Joseph Drambarean:

The big value that I see, especially watching customers at the Trovata level, but then also broadening the scope of this analysis is that even if you look at the use of ChatGPT for generating content, for example, for social media or generating content for writing in generally

Jeff Macke:

Generating content for financial reporting, for instance, when those companies report earnings, there’s only so many different things they’re going to say. They’re going to give you a few different numbers and you can actually have that.

Joseph Drambarean:

The common theme is boilerplate, right? It’s taking established practice, which is difficult to predict and giving you a shortcut almost from a creative perspective to get down to the brass tacks.

Jeff Macke:

Oh, not all, talk to people in the publishing industry. They’re outsourcing all that. If you’re writing a sports report, if you’re covering a World Series game, they scored three runs in the bottom of the ninth. You can tell that story 50 different ways, but ChatGPT is actually putting adjectives in there for you now. It’s almost impossible to tell the difference.

Joseph Drambarean:

The more that you can depend on it, which you can for these use cases, the boilerplate use cases, the more efficient you are at producing value in the high leverage tasks, right? Where you need to put a lot of time and effort into the thought capital of what you’re doing. And I think that’s where we’re seeing ChatGPT be a game changer, especially if you are in a role where you’re doing repetitive tasks, daily tasks where the same thing is happening every day, and you can use a predictable machine like this to give you that leverage. If I have the same report that I have to produce the same insight that I have to find and it takes me hours and now I can just put it through this machine and it does it and it does it without fail and it never gets tired, and I can ask it at any time of day and I can do it during coffee, I can do it while watching Netflix. I can do anything really.

Jeff Macke:

You can use it to order your coffee. That’s what we’re seeing. The first case uses that you see on a consumer level involve their drive-through, involve just talking to it, ordering fries from somebody. It’s always polite. They have no problem with accents. The speaker doesn’t matter the quality of it. It always comes out clear, concise, and they always upsell. It’s better at selling fast food and it replaces kind of drive-through workers right away. We’re going to see that on the finance side as well. I mean, it just seems intuitive. Brett, where do you see the next big lever coming specifically within the different functions of finance?

Brett Turner:

Yeah, I think that’s like we’re talking, there’s going to be, and I think it’s probably going to be a couple of basic reports. It starts to leverage in finance, but one of the things that’s on the enterprise side or the bigger the business gets, the more mature it gets. A lot of the things that are being done in finance start to get a lot more predictable. Doesn’t mean they’re easy to do, but they start to follow more of a defined boilerplate. Those are the kind of things that are just going to be so ripe we’re going through. And once there’s a level of, so you’ve got two things that are kind of counterbalancing. One is can it do the work? Can it do it accurately? And I think once there’s going to be a number of things that you’re, and some of our customers are already starting to use our gen ai, but when you leverage that, it’s almost like it’s a toy or you’re just kind of a little bit of a dress rehearsal.

You’re going to do it alongside maybe you doing it as well. And once you start to see it enough times to where the accuracy is there and you’re not seeing that deviation much, then the reliability starts to happen and there will be something along those lines that somebody speaks about and just has a little bit of that. It just impresses upon them that they literally just saved hours of time by doing this one thing. And once that goes, you’re talking about tipping point drivers because there’s a sort of best practice aspect to this too. It’s not like finance gets reinvented or that’s part of the downside. It’s not. Creativity doesn’t flourish. It can be really helpful if you’re trying to, from an analyst perspective, but oftentimes when it comes to your core stuff, it follows a certain rigor, it follows a certain level of predictability, and you’re trying to maximize accuracy.

It’s got to be accurate and minimizing the risk of, so when all of a sudden it comes out and you’re leveraging it to do a couple of those things, and then that becomes acceptable best practice, then everybody will instantly adopt. That’ll just become mainstream just like that. And I think we started seeing that with certain early cloud workloads, email, things like that. And then it just started to go and permeate the rest of business. And then when you look at cloud like AWS, now, it’s like this spend sprawl because it’s so easy to use AWS then everybody. So all of a sudden the finance is having to watch the bills of it spiraling. And I think in the same way, you’ll see sort of a, once that adoption starts to go, then it’s like, well, you’re running all this scenario analysis, but we just got a bill for it because it’s still pretty expensive to do that. We’re still not quite there yet. We’re just in these early phases, and especially in finance, you’re dealing with, these are professional risk managers. Many of them are here at AFP, right?

Jeff Macke:

Whisper voice. Some of these people are doing stuff that’s going to go away.

Brett Turner:

You’ve got to get by that level of rigor. But I think there’s another thing too that you see the dynamics at a conference like this. You see a younger generation, younger people, they’re more aptt to, there’s more trust in understanding how to turn the knobs and push the buttons or leverage technology in a way that, so you’ve got this great knowledge and yet more of this acceptance of leveraging new things. And so I think those are going to start to come together a little bit too. It’s going to be fascinating, but we’re clearly on all this just happening in the last…

Joseph Drambarean:

There’s a point that I want to pivot off of that’s interesting. It’s variation being a leverage point right now, one of the most fascinating use cases with Dolly or with ChatGPT is when you give it a prompt and then you say, now say it as a millennial or now say it from the perspective of Darth Vader, right? You can throw any variation at a basic prompt and it will find and intuitively discover the right way to give it back to you. Now, flip that to the finance world today. If you want any form of variation in a task that’s actually cognitive load, if you say, well, what if we did this same analysis, but from eight different perspectives, the analyst will say, do you realize what I’m going to have to do to do that? I’m going to have to spend the next six months.

I have to put all these spreadsheets together. I have to create custom models for every single one. Are you kidding me? I can’t do that by tomorrow. Well, imagine if all of that cognitive load was completely removed, and interestingly enough, the reduction of cognitive load in this case doesn’t have to be accurate because the variation is the decision point of should we even go down the road of looking at this angle? And a lot of the time you throw out the angle altogether because of the amount of effort it’ll take just to get to the answer. So imagine if you have that kind of leverage of not just eight variations, you could do a hundred, you could do 200, a thousand if you wanted to, and there’s no limit really to the creative capacity. What does that do for us as financial analysts? I think it completely changes your ability to be strategic.

Jeff Macke:

And are you seeing the ability to use these applications in an economically feasible way? As you say, Brett, it’s like cool toy and it seems fun when you’re playing with it on the internet. It’s actually really expensive to use this stuff on a one-off basis. The computing power required, is that the bottleneck? Is it just GPUs? What is actually going to be change as to make this more accessible to more people?

Brett Turner:

If it was mainstream and everybody’s using it and running all kinds of, as you’re analytics department and running all kinds of scenario analysis, it would be really, really expensive. But by the time there’s this mode of adoption, the costs are going to be coming down massively. It’s like the plasma TV when it comes down. Not too many people were buying that.

Jeff Macke:

The 42 inch $5,000 TV that was only four inches thick, although that just crushed TV cabinets as an industry. If you’re making TV cabinets, man, Joseph’s looking at me, he’s never even seen one. Before all the TVs were flat panel olds like us, we had cabinets.Let’s put on our futurist hats. Let’s put on our Tony Stark hats. What’s the risk we don’t see coming, but on a societal level, blowing up past just financial professionals, but what kind of risks are we running here? The classic example already just being military drones where you kind of set ’em free and they turn on themselves and eliminating the people trying to control the drones becomes a rational solution. What are we playing with you guys mind? People who play with their stuff all day?Joseph as someone who develops it. What’s the biggest risk that you think people are maybe walking on a tightrope they don’t understand?

Joseph Drambarean:

It’s common across every industry. You just gave the military example, it would be the same exact risk. It’s bias mitigation. It’s how can we ensure that given autonomy, this system will not be going down natural roads of bias that will lead large groups of people to believe that they’re seeing accurate information. But in fact, it is from the lens of a bias that has been established because of an unknown amount of algorithms playing out different scenarios and evolving over time. And I think that’s the big fear that we have with the military side we have with the finance side of the house. Even when you think through the perspective of trading on the trade floor, this is one of the biggest fears. If an autonomous system is manipulating mass market, what does that mean, right? If it’s doing things by itself, the same conclusions can be drawn. I think from an analysis perspective, and also when you think about a large of these corporations, the largest use case that will come will be handling money.

The next step is once you trust these systems to make judgment calls from an analytics perspective, the next judgment call is, well, can it also balance our funds, make the right decision strategically for us to move funds in various places? And that’s where bias becomes a risk, right? What is it doing? Why is it doing what it’s doing?

Jeff Macke:

If you draw the check on that though, from the trading example, you end up with, you’re front running trades, you’re executing things that you don’t understand, you’re taking risks that you’re able to get much closer to fat-tail risk events to running it off a cliff without really knowing it just based on some rational assumptions of what’s happened in the past. And so you, you’re inputting probability at a base level that then plays out as it just becomes part of the infrastructure.

Brett Turner:

Well, I think an example what Joseph is saying too, and this is still down the road, there’s going to be a lot of basic things, and then all of a sudden you’re realizing that comfort level goes up. You’re going to say, well, there’s this particular wire or even investing our excess cash we want to get, and we got to send all these wires to do that across all our accounts. Wouldn’t it be nice if we turn that over just to analyze the amounts we have across all of our accounts that are in excess by thresholds that we sit or we set, and then go ahead and even just transact those for us at the end of the day so we’re not having to go through all those templates and manage that. And it’s a pretty repetitive task. And so then you can say, well, that’s happening. It’s automated. AI is doing that for you. And then all of a sudden it’s like something happened where they decided to chase yield. Like we found out, oh yeah, there was this one high yield, some junk bond that was butler’s just like, yeah, but instead of getting 4% or three and a half percent I could chase after this 9%. To me, it seemed like a really, really good idea. The calculus said, put all our money in that. Well, obviously that can’t happen.

Joseph Drambarean:

Well look at the SVB examples. That wasn’t a junk bond. That was government bonds. They were the most secure bonds that you could have, it was just a misallocation, it was a risky portfolio considering the event, considering the time that we were in. And I think that is the bias that you always run the risk of with these autonomous models is that they might be making the right call in the moment with the data that it has at its disposal. Right?

Jeff Macke:

And you don’t find out that fire’s cool, we’re playing with fire. It actually works out really well. Prometheus wasn’t an idiot. He brought it down and it was nice and it’s useful and you can cook your food and oops, every once in a while it just kind of blows things up. It seems to be a risking take. Joseph, I want to finish with in the last eight months, nine months since the collapse of SVB and since the release of the mainstreaming, if you will, of the open AI, you talk about security being the first thing that people are trying to do now that they trust a little more, how they’re feeling a little more comfortable with the security, they’re expanding, how they’re using it, what’s going to be next? Help me kind of know where the puck is going is Wayne Gretzky used to say.

Joseph Drambarean:

Yeah, there’s a concept that is emerging that is called multimodal generative AI. So as if we didn’t have enough of these acronyms and phrases. What it’s saying is that what if AI had exposure to many surfaces, the internet, for example, and it could actually just search the internet like a human could. What if AI could watch video? What if AI could look at photos? What if AI had access to your email? Right?

So this multimodal concept is saying if we expand the surface area of things, of data, points of variables that AI could be using with the exact same generative finance or generative tactics that it was using before, could it be smarter? Could it be more real time? Could it be more predictive? Could it be better at finding things? And that’s where we’re seeing the latest innovation. So just a few weeks ago, OpenAI released with ChatGPT four, the ability to feed it, photos and videos, the ability for ChatGPT four to do internet searches. Bing has been using this for a while in partnership with OpenAI, and I think that’s where things get super interesting because it’s like, to use a Marvel analogy, it’s like the Marvel universe has expanded now to the multiverse. You have all of these different things that it could be looking at, and I think that’s where these concerns that we’re bringing up bias, where’s it getting its information? Is it leading it in a certain direction? That’s where we have to have our guard up.

Jeff Macke:

We live in an age of miracles. Horrible, terrifying, kind of awesome miracles, and you guys are right on the forefront. Hey, thanks a lot you guys. Thanks for letting me sit and answer some questions.

Brett Turner:

Awesome, thanks.