Jeff Shen – Beyond the Buzz - What AI Really Means for Investing

00:00:07 [Jeff Shen]
I think the breadth and also the depth of research agenda, how much we're pushing on future output signals or ideas, that's oftentimes foundational to our future success.
00:00:22 [Vincent Weber]
Welcome to Resonanz Spotlight, the podcast where we explore the people and ideas shaping the future of investing. I'm your host, Vincent, and today I'm truly honored to be joined by Dr. Jeff Shen. co-chief investment officer and co-head of systematic equities at BlackRock. Jeff has been at the forefront of quantitative investing for more than two decades, pioneering a thoughtful blend of human judgment and machine intelligence. In this conversation, we'll explore his personal journey from the early days of quant to the rise of AI in asset management. We'll also dive into how to lead high-performing research teams, how to manage innovation at scale, and what trends he sees shaping the next generation of systematic investing. Jeff, welcome to the show. It's a real pleasure to have you here.
00:01:15 [Jeff Shen]
Thank you so much, Vincent. It's great to be here. Very honored to be on the podcast with you.
00:01:22 [Vincent Weber]
Thanks. Let's start at the beginning. Your career began in 1997 at J.P. Morgan. where you focused on global macro investment and asset allocation research. Later, at Barclays Global Investors, you led Asia-Pacific and emerging market active equities before its merger with BlackRock in 2009. How did this formative experience influence your thinking and shape your approach to systematic investing?
00:01:51 [Jeff Shen]
Yeah, that's a very good question. I mean, I'll say that, right around that time, before I went to JP Morgan, I was a student for many years, sort of eventually ended with a PhD in finance from NYU. So New York City, 1997, and then started working at J.P. Morgan. So I say that it was to a certain extent, if you will, in analogy, it's a little bit of a hammer and that's looking for nails. So I have certain set of skills in quantitative systematic. investment and then really was looking for the applications in the real world, so-called. And JP Morgan, as a management, the division that I was in was certainly a perfect environment for that because it has many different investment strategies, many different approaches for investment.
So as a fresh PhD coming into the industry, that was certainly to your point, very formative. And I think, around that time, I would say that from a macro investment perspective, it was also, I came to the macro team that was reasonably discretionary, fundamental driven, and certainly was trying to make both longer term strategic asset allocation, but it was also trying to look at some of the tactical asset allocation. And I think with my systematic quantitative training, I think when I was at JPMorgan, I certainly saw the opportunity of using some of the quantitative tools to help to enhance the discretionary macro investment. So I did that in New York and also worked in JPMorgan in London. And that was a great exposure across many different parts of the firm, different parts of the strategies.
So I think it's almost like in the United States, when you go to a college, you have a two-year, if you will, almost like a liberal arts training before you decide on your major. So that was a sort of a very broad-based training and very good exposure. Then when I left J.P. Morgan Asset Management to join Barclays Global Investors in 2004, I think part of the reason is Barclays Global Investors at the time, certainly had one of the best quantitative systematic investment platforms in the world. Lots of professors, lots of PhDs. Some of the folks even have office hours posting on their office doors to answer different questions. It was a very academic-oriented environment.
And I think that was certainly from a culture, from a environment, from a colleague's perspective, I certainly found a lot of people who spoke similar language, a lot of people who are very similar-minded in terms of how we think about the market. And I think I delved a bit deeper into the Asia-Pacific and emerging market because that was an area that when we were initially doing some of the systematic quantitative investment in security selection, that part was lacking. And that part was certainly emerging, if you will. So I spent a lot of time focusing on that and, really try to grow that investment capability.
And that was very formative in the sense that I think it really made me to understand, even though you're using a lot of quantitative systematic tools, it's very important to understand specific market contexts. clearly Asia-Pac and emerging market is a very evolving, complex market. So I think in that sense, it is actually complicated enough. And also some of the local characteristics, whether it's the importance of government policy or whether it's retail investors moving around the market quite a bit. So there are some local contexts that actually make it quite fascinating to delve into. So I'll say that, yeah, I mean, I think JP Morgan was very good in terms of giving up that broad base.
And the BGI was certainly very good in terms of understanding the application of the quantitative tools with a special awareness for the context. And fast forward today, I've been with the same group for about 21 years. And, clearly it was BlackRock proper for about 16 years.
00:06:35 [Vincent Weber]
Thanks. That's very interesting, especially because. I've got a feeling that right now we are really living through a transformative moment. You have big data and generative AI, which is rolling across industry, including our own. So how do you see AI reshaping the landscape of systematic investing, and especially, I would say, for you here at BlackRock today?
00:07:02 [Jeff Shen]
Yeah, I mean, I think obviously that's, I talked to you a little bit about my JP Morgan days, BGI days, and I think the last 16 plus years at BlackRock proper, I would say that it's certainly been sort of coinciding with our heavy investment and interest and development on using artificial intelligence for systematic investment. So while today, to your point, AI is a very popular topic, I would say that we've actually been using AI, leveraging AI for the last 15, 16 years. So it's been not necessarily something new, but it's certainly something that we have very strong confidence and strong conviction in. And I think if I delve one level deeper, I'll highlight three things on AI.
I would say that AI, the first dimension of AI, I would like to think about it as big data. And big data can certainly be thought about as using non-traditional unstructured data to help systematic quantitative investment. And so the data examples can be various different texts. It could be a conference call transcript, could be a broker report, could be financial news, anything that's actually text-oriented. It could be unstructured data in the form of a satellite image. or Wi-Fi beacon information, or credit card transactions. It could also be a form of data that's really trying to, high velocity, high volume, very noisy, things like social media information that could also be helpful to dissect a bit of how the retail investors are thinking about.
I think the first dimension I'll say that is really big data, alternative data, lots of different data. I think the second dimension that we like to think about application of AI is really what I call as a big model. The big model is really that I think the models are getting bigger and bigger and more and more sophisticated. While everybody is probably quite excited about large language model, which is rightly so, I think a lot of it has to do with it's a large language model. The fact it's a very large or big model makes a big difference. So I think, if you think about maybe some of the audience here has done linear regression before or in your education. And so that's a very simple linear model.
Fast forward today, the model can certainly be, billions, billions, if not trillions parameters, neural network, very sophisticated, very, large. So machine learning model. A lot of this evolution has really evolved in to go from simple models to much more complex, much bigger models. And I think the bigger models that actually use lots of computation power turn out to be very effective in many different domains. And I don't think that investment of finance is actually special. I think it's going to see a lot of the efficacy from the big models. Last but not least is that I think I call it as a big crowd. I think when I say big crowd, what I really mean is that it means that when you use artificial intelligence, I think it's really becoming a scale game.
So it's actually allowing for scale to not only to get data, get a large model, but it's also allow for different disciplines and different strategies. and different investment horizons, all blending in together. Vincent, I know that you've certainly been at the forefront of thinking about some of these multi-strategy, high alpha strategies. And I think that's how we're actually thinking about when you have multiple things going, when you have that scalability, that can potentially drive much more consistency and can potentially drive more differentiated returns. So I'll say that that's at the big picture level on the- how AI is shaping systematic investment. In my mind, it's really, it's about big data, it's about big model, but it's also about big crowd, try to use multiple strategies, try to leverage the scale.
00:11:32 [Vincent Weber]
Well, very interesting. And could you share some, one specific example, how AI tools are being actually used in your investment process?
00:11:45 [Jeff Shen]
Yeah, I mean, I think I'll say that all of us probably have at various different moments read broker reports. So reports that come in from sell side and, probably have like 15 of those sitting in your inbox. And I think we've been using essentially natural language processing to process these broker reports over the last 15 years. And, when we first started 15 years ago, the methodology was very simple. There's this machine learning algorithm called bag of words, which essentially you're looking at a document. You are trying to figure out what is a positive word, what is a negative word from a particular document. And then you try to add the positive words and negative words together to figure out the ratio of the positive versus negative and then figure out the sentiment. So that was.
the state of the art back 15 years ago. And we've actually experimented with that 15 years ago. And, that particular natural language processing reading broker reports has been evolving over time. And we actually now have seven versions of it. And clearly the version as of late has a whole lot to do with using large language model to essentially read a broker text and try to really gain a much more nuanced and deeper understanding of what the brokers each report are saying for a particular company. And so I think we've been using that. And my joke on this is that we like to hire PhDs who don't like to read. So they will be using natural language processing algorithm to read a lot of these things. So I think that particular example has been.
around for us for a long time. And I think the interesting thing is actually the performance coming from that has been very strong and also very differentiating, especially compared to the generic analytics revision, where people are just looking for earning upgrade or downgrade in a very generic sense. So I think that's, I would say that's one successful example of using the AI tools where you can do much, much better than the generic tool of using analytics revision. by leveraging natural language processing, and especially as of late, this kind of large language model.
00:14:16 [Vincent Weber]
Right. Thank you, Jeff. Jeff, probably you're wearing many hats. You're a researcher, but you're also a manager of people. And so you're leading a team. You're responsible for managing billions through quant models and models that must adapt and evolve with the market. So how do you balance the push for innovation with operational discipline required in large-scale investment business?
00:14:46 [Jeff Shen]
Yeah, I mean, I think it's definitely a team sport. To your point, there are many dimensionalities that we need to be excellent at. So certainly this is not a star portfolio manager type of culture, type of setup. I think we really think about systematic quantitative investment as a team sports and everybody's going to play their role. And I think, to your point on this balance between innovation versus operational excellence, I'll say that it's really, I mean, the way that I think about it is actually on one side of, try to be world-class and excellent has to do with the try to evolve and innovate on how we forecast future returns. So if you will, this is the alpha forecast. And that's a game of innovation.
I think the breadth and also the depth of research agenda, how much we're pushing on, future alpha signals or ideas, that's oftentimes foundational to our future success. So that we are, we have lots of researchers. I mean, it's a big team. It's, in totality. On the systematic active equity side, we have about 150 people. So there are a lot of people who are focusing on research. At the same time, to your point, we do have billions of assets. So there are actually also portfolio managers and strategists, people who are also really looking at making sure that risk is well taken care of. And we're looking at the different costs in a very extent and also measure some of the market impact in the exposed setting.
And there is this concept of transfer coefficient in the sense of how much of your insights can be actually translated into the actual portfolio. So from an operational discipline and operational excellence perspective, there's a lot of focus as well. So I think, yeah, I mean, I think I'll say that the answer to your question is really, I think I'm trying to be not only a researcher, a portfolio manager, a person who is also trying to help to think about strategic direction of the group, but ultimately, it is also having a lot of people who are much smarter than me to be able to do work and then contribute, all of us contribute together as a team.
00:17:17 [Vincent Weber]
All right. Just mentioned the concept of transfer coefficient. I mean, this is a This is a concept nicely discussed in the original Grinnell-Kahn book. So is it still a part of this historical legacy in your team? Yeah, yeah, yeah.
00:17:35 [Jeff Shen]
Very much so. I'm so happy that you mentioned Grinnell and Kahn. I mean, clearly, they've written the book, in my mind, on active portfolio management. And Richard is happily retired. Ron Kang is still our global head of research. He's been with us for 28, 29 years. And I've only been with the group for 21 years. And so I think, yeah, to your point on transfer coefficient, information coefficient, and the concept of breadth, lots of the fundamental, a lot of actual management is very much well alive in terms of our investment philosophy and investment process.
00:18:18 [Vincent Weber]
Right. That's great. I think that's also a book which was massively ahead of its time. Right. And remember, the first time I read that book, I said, oh, this is totally useless because I was expecting, OK, what, what for about to read about forecasting factors and what variable had me predict things. And it took me many years, many years, where randomly, came the situation where to take another look at this book, where I understand, what the value in this book, where in fact this book was. disclosing a lot of secret recipes, which you wouldn't see at the first rate.
00:18:55 [Jeff Shen]
You use the word recipe. I think it's a very important word. It is actually a very good recipe for active portfolio management. Clearly, different people may use different ingredients, to come up with different outcome. And we certainly do look at, that. But I think the framework of thinking is timeless. and extraordinarily transformative.
00:19:19 [Vincent Weber]
Absolutely. Coming back to people's management. So what have you learned over the years about, leading, leading Quentin in such a dynamic environment?
00:19:31 [Jeff Shen]
Yeah, I mean, I think I'll say that it's, a couple of things to think about on the people front, because I think sometimes when people think about quantitative systematic management, People like to think about the machines that are doing the work and the humans are not that important. It's about formulas. It's about data. It's about algorithms. That certainly is true to a certain extent, but it's really, I'll say that probably only half or maybe even less than half of the story. Ultimately, I think for a world-class systematic quantitative investment platform, it's actually a big chunk of it has to do with the people. So I think it's certainly something that as a CIO for the group, I spend a lot of time to think about that.
And I think we are looking, certainly try to put a team together that actually has a lot of diversified skill sets. I think the one big evolution that I'll say that over the last 20 years, we've actually evolved is when I joined the group, I think that we had a lot of economists. people who have finance and accounting background. And I'll say that there's not much else. And fast forward today, I'll say that the team is much more diverse in terms of discipline. And I think the big transformation has been really anchored around getting more people who have computer science, engineering, just STEM majors in general. and really try to get more of engineering technology and the compute culture into the group. So I think that's been a big transformation.
And I think it's, I think that's when we talk about artificial intelligence and machine learning, a lot of times it's not really about the talk. It's actually really about people who have world-class expertise in this field. We've actually worked with a few professors at Stanford University, which is pretty close by near our office in San Francisco. And they've also been invaluable to add a bit of a foundational support on how to think about the application of AI and machine learning for investments. I think that's probably having different disciplines represented in the group is extraordinarily important. I think the second one, I think you also mentioned that the world is changing fast and it's a very dynamic environment.
So I think we want to hire a diverse set of talent who are actually very smart, especially on machine learning or AI applications. At the same time, I think we are also always looking for people who can hustle because markets are moving fast. So to a great extent, it's not about, and I think that's probably one big evolution for us over the last. 20 plus years is that I think when I joined, I told you about, the professors, the PhDs, the academic environment, which we still preserve a lot of it. At the same time, I think the other ingredient that we added to the recipe, if you will, has to do with adding this factor of hustle.
It's sometimes, market is changing and you got to really be a bit more practical and be a bit more nimble and you got to have that extra gear. to make sure that we deal with either on a risk management front or trying to come up with new ideas. So I think I'll say that's probably the other element we've actually added. So I think we want to have good people, people who have good characters and people who can work with one another. The collaboration aspect of it is also extraordinarily important. And then we also want to have people who can hustle in a pretty dynamic marketing environment.
00:23:33 [Vincent Weber]
Thanks. So looking ahead and beyond the AI burst and hype, what are the emerging trends or shifts you believe will define the future of systematic investing?
00:23:45 [Jeff Shen]
I think, yeah, I mean, to your point, right? I mean, I think AI is on one side of it is I think it's overhyped. On the other side of it is I think it's also underhyped. And the reason I say that is if you look at a classic textbook by Peter Norvig on artificial intelligence. There are 28 chapters in that book. And there are two chapters on natural language processing. And if you think about large language model, it essentially belongs to using machine learning for natural language processing. So all the hype has been around that two chapters. And it's great. There's some major breakthroughs. But I think there's a if you ask people today, what do about AI?
I think a big majority of the people are really thinking about a large language model equate that to AI. But it turned out to be that when you look at a textbook on artificial intelligence, there are 26 other chapters that not too many people are really talking about. And I think that's probably the portion that is actually undervalued or underhyped. what about reinforcement learning? What about deep learning? What about online learning? So there are plenty of things that actually are in other chapters of this book. Or if you go to a world sort of AI conference, you're going to realize that whether it's multimodal learning or, there are a lot more additional topics that is actually happening that I think is actually a bit underhyped.
When you ask me what's the future, I do think that this, how AI is going to transform investments, systematic investment in particular, I think is extraordinarily significant. At the same time, I just want to say it's big umbrella AI. It's not AI as defined as just only large language model, but it's really to think about the full stack of AI that can potentially transform systematic investment. And there are many. many exciting applications.
00:25:59 [Vincent Weber]
Absolutely. Thanks. And what risk or blind spot should professional investors be paying closer attention to over the next five to 10 years?
00:26:10 [Jeff Shen]
Yeah, I mean, I think I would say that in general, I do think that when you think about the professional investors, I mean, I think I would say that, a couple of things. that I think people may want to watch out for. One is, clearly we are entering into a regime and also an era that there are plenty of policy volatilities and also policy uncertainties. We've seen a bit of that on a year-to-day basis. And at the same time, I think some of these policy uncertainties and the policy volatilities are here to stay. So how do you evolve an investment process to think about some of the policy volatilities? And how is that driving security selection? How is that driving asset allocation? How is that driving overall some of the macro assets?
So I think that could be an emerging risk, emerging a blind spot that we may want to think about. So I think that's probably number one. Number two is also, I do think that investment management in general is getting to be more and more professionalized, more and more institutionalized, which is in itself an interesting pattern. I think it's also important in light of that to think about the potential crowding, potential breaks in the, if you will, the financial cracks in the system, because there could be potential crowding and potential crowded positions from a risk perspective that a professional investor needs to think about. Last but not the least, I'll say that we talk quite a bit about artificial intelligence and how that's going to be transformative for investment on a forward-looking 5-10 year basis.
I think in light of that, I do think that transparency and also a deep understanding of the AI stack that people use is also quite important. It's certainly okay to use a large language model, but do you really understand what's going on behind the surface of a large language model? It's certainly great to use reinforcement learning algorithm, but do you really understand what exactly the penalty terms are, that the learning objectives are?
So I think that's probably, I wouldn't necessarily call it, sort of a risk per se, but it could potentially be a blind spot and it could be, from an investment process and risk management perspective, it's certainly a new, new thing that I think the investors need to dig a little bit deeper and to make sure that they understand both the opportunities, but also potentially the risks associated with it.
00:29:10 [Vincent Weber]
Right. Jeff, thank you so much for sharing your insight, your journey, and your perspective. It's been a fascinating look at the evolution of quant and a reminder that being systematic is as much about mindset as it is about models. To our listener, if today's episode made you rethink what's possible in data-driven investing, then we've done our job. You can find more insight and resources at resonancecapital.com and don't forget to subscribe to stay up to date with future episodes. I'm Vincent Weber and this was the Resonanz Spotlight. See you next time.
00:29:43 [Jeff Shen]
Thank you, Vincent.