Delivering Global Alpha

Key Takeaways

I'm going to start our conversation with AI, but I want to tackle it from from two different perspectives because you come at investing from two different ways, two different >> styles. So, you're a quantitative investor. Um, data science is always critical >> to how you make your investment decisions. You hired your first PhD later in in AI back in 2008. >> That's right. to help build out your strategies. You now I read said you needed a team of quote feature engineers. What is that? >> Right. So, so if you think about uh the the availability of data and the ability of AI and machine learning to turn any bit of information into useful data, whether it's imagery, whether it's text, sound, unforatted text. So back in the day, a new data set would be a new time series. Time series of prices, time series of volumes traded, but it was a time series. Nowadays, a data set is anything. A collection of words, a collection of sounds, a collection of images, uh, consumer data. So while a time series, everybody knows how to look at a time series. They plot a chart and the the horizontal axis is time. these less the these more amorphous data sets they they they have to be explored what are the features what are the variables that I'm looking at in this data set and so there is a discipline now that we call feature engineering look at this data set what are the horizontal axis that you're going to inspect the data on what what are the relevant parameters to characterize the data set and and so we do have a a research area within our company that is Traditionally research for us is alpha making. So in the last few years we actually formed a research team that was all about data. They're not alpha makers. They're just people that treat and interpret and select the data. >> But less than five years ago, these this group was alpha makers. >> Correct. >> And that just shows you how quickly this is changing. >> Yeah, that's right. That's right. I read where you said quote nowadays the data is so much greater in volume and so specific that it no longer makes sense for it to be in their domain [clears throat] speaking of your what was in the domain of your investment staff has now morphed to these feature engineers. >> Yes that's right. So so there is a a research activity that that is research it is creative. It is innovative. It it requires mathematical skills but it preempts the alpha making. It it's about understanding the data and exposing like they talk about h what elements have you exposed of this data set you know so you see it's you would never ask that question with a time series right a time series of prices a series of prices versus time you know there's time and there are prices and off you go but as I say the new world the data is such a broad concept now that that data sets need a bit of work before they can be used for for making. >> So if Absana if I if I called you a more traditional investor uh is that okay? >> Absolutely. >> Oh but these guys but no but these guys also use their data really well. >> Sure. But how how do how do you see the AI story playing out in from from your lens? >> So at Rock Creek we do two things. One is we invest in early stage uh companies particularly in energy innovation and then we invest as like the outsourced chief investment officer across lots of funds. So we have two sets of data sets. We were early investors in Y combinator when Sam Alman was there and started open AI. So we were fortunate in that way but since we started Rock Creek we have had a big about a third of the team when we started Rock Creek was computer scientists and quant people and what later said is super important because we've used this data in a huge way and today the way we can with AI um basically be so much more efficient with ingesting the data making it more structured making it uh so that we can make together decisions on our investments that has changed dramatically in the last 12 months in even more in the last 6 months. So uh >> one third of your staff >> absolutely computer scientists it's today it's changed because maybe about a quarter is and we just actually launched a group called um serious innovation lab literally on November one that has information on several thousand funds and is looking at what signals you can pick from that but also you can do your first due diligence all your documents everything is in one filing cabinet >> so it makes our work very different than it was 2 years ago. Are you >> I'm telling you, Scott, everybody's a quant now. >> Everybody's a quant. >> So true. That is so true >> because because the world I remember the first time I heard the phrase datadriven decision making and I thought to myself, is there any other kind? And then I thought, wake up later. Of course, there's only the other kind. People make decisions with emotions, not with data. >> Is it fair to say that that AI in some ways has democratized being a quantitative investor? >> Absolutely. Absolutely. >> Because we have access to data in ways and we can have it processed in ways that we never had before and we can use it to our advantage as we're trying to generate outcomes. >> Exactly. And and AI will help you even if you're not necessarily that familiar with anal. You've got to be a little bit careful, right? So like I I treat the the AI tools a bit like a keen interns, you know, like they're bit keen. So you've got to tell them exactly what bit of analysis you want them to do for you and check their numbers a bit. But in general, people with much less mathematical training can use the tools to analyze data and to to interpret data more. >> Do you ladies think we we'll get to a point and and maybe in the the near future where you'll you'll use learning algorithms to actually pick investments? Will will we get to that point A? How how you see that? I think today the way we see it is you will be humans using the algorithms to make better decisions versus the algorithm you know making a decision. things could change over time but at this point what we can do particularly in uh in less quantitative areas you do need the human and without that you can't really do your work where you don't need need us and the AI can do a much better job is what we talked about getting the data cleaning the data making it in a format we used to have you know 20 30 people just to do that you don't need that skill you need very different skills but that decision- making and judgement ment I don't think that's going away >> later do you agree >> yeah look I the the the mathematically the the big problem is that financial data is actually quite sparse right and quite sparse and there's a lot of randomness in it and and those two things make it very hard you know it's not your traditional deterministic problem like if you think about the self-driven car what is random when you get into a car and you ask a car to take you from here to JFK. What is random? You know the roads, you know, you've mapped them and and even if you haven't mapped them, if they've changed a little bit, the the car has cameras and and and so that's a mostly a deterministic problem. The the the the financial data contains a lot of randomness and that makes the decision making a bit difficult to automate.