Are we ready for a hedge fund world where machine learning guides trading decisions and data scientists outnumber traditional traders? In this Tech Talks Daily Podcast episode, I sit down with Gary Collier, CTO of Man Group, to find out.
Man Group traces its roots back more than two centuries to its days supplying rum, yet it now stands at the forefront of AI-driven finance. Gary explains how the firm's open-source platforms, Alpha and Rosa, power everything from signal generation to automated execution, all while keeping a firm grip on the art of human oversight.
I found it intriguing how their internal ManGPT system used by over half the organization proves that large language models and neural networks aren't just hyped. They're part of daily operations in purely systematic strategies and supporting discretionary portfolio managers who want real-time insights from a sea of unstructured data. We also explore ArcticDB, Man Group's open-source data science database, enabling it to process billions of rows quickly. Gary argues that without this kind of data infrastructure, AI research stalls under the weight of time-consuming manual tasks.
Throughout our conversation, Gary shares how his hobbyist coder and physicist background influences a data-first culture. It was fascinating to hear how he believes AI could soon disrupt creative aspects of quantitative research, allowing advanced models to generate strategy ideas rather than simply refining existing ones. At the same time, he underscores the importance of transparency. In a world where billions in assets are at stake, pure black-box automation doesn't cut it, and teams need to explain how trades are executed and why the data looks the way it does.
From front-office analytics to deep research projects, it's clear that AI pulses through Man Group's veins. Still, Gary reminds me that people remain vital for framing the right questions and deciding when to trust the outputs. Will AI become the main engine of future hedge fund strategies, or will human ingenuity continue to guide our biggest calls on the market? And how will your organization embrace the growing tide of machine-led innovation?
[00:00:04] How is AI reshaping the future of asset management? Can machine learning not only streamline operations but also disrupt the creative side of quantitative research? And how do firms balance cutting-edge automation with the human expertise needed for complex financial decisions?
[00:00:25] These are just a few of the things that my guest Gary Collier, Chief Technology Officer at Man Group, is going to help me with today. And Man Group is a firm that has been at the intersection of finance and technology for decades. From its origins as a rum supplier over 250 years ago, to becoming one of the world's largest publicly traded hedge fund managers, Man Group has always embraced transformation.
[00:00:53] And with 30% of its workforce now tech-focused, Man Group is leading the charge in leveraging AI and data science across its entire investment process. Whether that be from data onboarding and risk analysis to trade execution and strategy development.
[00:01:13] So my guest today is going to be showing how proprietary platforms like Alpha, which powers their front office, and Rosa, their operating platform, are collectively driving efficient innovation and giving Man Group a competitive edge. And I think this is so important. We talk about the ROI and measurable difference of technology and what it can make.
[00:01:34] And today we'll find out all about that and how AI models, including neural networks, natural language processing and reinforcement learning, are ultimately shaping trading decisions. And with machine learning now powering 25% of the firm's trading models, there's a lot of exciting things to talk about. So can AI truly revolutionize the creative aspects of an investment strategy?
[00:02:01] And how do firms like Man Group ensure AI explainability and data quality while pushing the boundaries of innovation? Well, it's time to find out. Enough rambling and scene setting for me. Let's get Gary onto the podcast now to talk about all this and much more. So a massive warm welcome to the show. Can I tell everyone listening a little about who you are and what you do? Yeah, I can indeed. And great to be here. Thanks for the welcome.
[00:02:28] So Gary Collier, CTO of Man Group. That's the short answer. But I think the slightly longer answer to that involves a bit of history about the firm. So Man Group were a large global tech powered active investment management. The tech power things are very important there. We've got a long, I think quite interesting history. We started life in London over 250 years ago. We started off as a supplier of rum to the Royal Navy, expanded into commods trading.
[00:02:56] Then in the mid 90s, we acquired a firm called AHL and they were one of the systematic quant pioneers. So that brought really data and technology front and center in investment decision making. Of course, subsequently, we've expanded the portfolio there. We've built out different investment businesses. We've acquired other businesses. We now did discretionary investing. We built solutions.
[00:03:21] So against the backdrop of all of that, I joined the firm in 2001 as an individual contributor, as an engineer, building out quant systems for AHL. They were pretty small at the time, less than 10 people, I think. And yeah, I guess I've been part of the journey, helped the firm grow, helped make sure that that initial tech DNA that AHL brought to the firm continued to course through it.
[00:03:50] And lots of different threads of tech, of course, have come and gone and have been woven together. And that led to the role I'm in there a couple of years ago being asked by a new CEO to bring all tech together into one single unit. So that's me. That's my role. And we're at 30% of the headcount of the firm is tech. What a great story. And I did a little research on the company before you came on the podcast.
[00:04:17] And I completely missed that history around RUM as well. I love that journey from RUM to technology. And of course, fast forward to present day, you are CTO of Man Group. You oversee technology and data science across the firm. So I've got to ask, there's so much hype and talk around AI at the moment. How is AI currently being integrated into your trading strategies? And what kind of impact is it having on hedge fund investing?
[00:04:45] Because we don't talk about this stuff enough, but it is just about everywhere at the moment. So what are you seeing? Yeah, well, you're right. There's loads of hype about it, of course. And if we looked at all that hype, we might think that the AI in investment management is something like no, but of course it's not. And we started to explore the use of AI for investment management in Man in the noughties. So a long time ago, and I'm not sure many people remember this. I can remember.
[00:05:15] It was not really that fashionable at the time. We'd come off the back, I think, of one of the AI winters. But really starting around 2010, we began investing an awful lot in building at the time what were quite big compute clusters, investing more in data. And so we've been exploring and using AI in many different parts of the investment process for some time.
[00:05:39] Initially with simple stuff, cane nearest neighbors type models, nonlinear relationships between market signals. We explored recurrent neural nets for time series analysis. Of course, time series and analysis of them is a big part of finance and financial data science. And then NLP for dealing with documents and text, reinforcement learning for better execution. So we've been doing all of this stuff for some time.
[00:06:06] And that decade, 2010 to 2020, saw an awful lot of activity. But of course, it didn't quite have the hype that got generated when ChatGPT and GenAI came along. And we were chatting seconds ago about where we were located and said, I'm sitting here in Boston in the seaports. And I'm sitting here in the offices of Manumeric. That's one of our quant businesses.
[00:06:34] And about a quarter of the models that are running now generating trading signals, investment signals for this business are based on machine learning techniques. So this really is, as a genre, nothing really new to what we do. But to fast forward to current times and I guess all of the hype that's around GenAI and how that's affecting things.
[00:07:02] And that's been quite interesting because the last decade or so before this, you could rely on new tech being primarily of interest to the quant businesses, right? They were the ones on the cutting edge of tech. And GenAI changed that quite substantially. Really, for the first time, the discretionary investment businesses and other parts of the firm as well were asking, how can this tech help us? So that's been quite novel.
[00:07:31] That's been a little bit different. So we're using GenAI techniques, not just in, say, the systematic business, but also to help discretionary portfolio managers better understand the information sources that they've got and hopefully make better decisions off the back of that.
[00:07:52] We've also built out to think because a lot of firms have our own internal, broadly accessible GenAI tool called ManGPT and what that is. That's essentially a safe and audited means for anybody in the firm to access one of a range of different LLMs. And that's pleasing to see that that's actively in use now by over half the firm on a regular basis.
[00:08:17] Yes. And before you join me on the podcast, I was reading that you're a big proponent of technical platforms. And at Man, you've invested a lot in both your front office with the Alpha platform and in something called Rosa as well, which is, I believe, your proprietary operating platform. But I think right now there is a real big focus, especially this year on ROI and the value out of any AI project, in fact, any tech project.
[00:08:43] So can you tell me a bit more about how these platforms make Man Group's approach a little bit different, stand out from the crowd and how they're ultimately leveraging AI and data science to make a real tangible difference? Yeah, you're right. I've been a big proponent of platforms for a long time. And we all, I guess, read lots of things or influenced by lots of things over the course of our career.
[00:09:08] And some things really lodge into our consciousness and then direct subsequent decision making. Like, I can't remember when or where I read it. It was a long time ago. But the notion that technical platforms, when done well, what do they do? They democratize the ability to innovate. But at the same time, they centralize and make efficient all of the common heavy lifting.
[00:09:35] And that notion is something that has really stuck with me for a long time. And that notion of efficient innovation and using platforms to drive, technical platforms to drive efficient innovation, is very much part of the heart of how I view tech and ethos of tech within the firm. Of course, there's always, I suppose, when you're deciding what to build, what to buy, this buy versus build debate,
[00:10:03] it's one of those perennial tech questions that never goes away, never seemingly gets out of fashion. Now, a large part of the approach to building platforms at Man, particularly in the front office, what we call our alpha platform, it's been an open source first approach. So I think that combination of bringing together, I think at the last count, there's somewhere between 1,500 and 2,000 open source packages that we bring together into a coherent platform whole for the front office.
[00:10:33] But that combination of open source and smart engineers that can wire things up, add your own USP, really drives home that ability to innovate, to do so efficiently. But also, I think importantly, and this perhaps is something a little bit overlooked, it gives you that agility as well. And in this whole buy versus build debate, I think it's often worth considering paying that extra premium
[00:11:03] if it buys you some agility. And the balance between those three things, innovation, agility, and efficiency, I guess at an abstract level, are what drives the whole platform debate and how we build tech. So if we pull all of this together, what do we end up with? We've got this hugely rich and powerful Python-based front office platform that's built on a lot of open source, but 10 million, 20 million lines of our own code as well.
[00:11:33] And we've got an operating platform, Rosa, that you mentioned minutes or so ago, that very much in the listed space allows us to trade pretty much anything and gives us the ability to combine all of those pieces from the front office that are built using the Python platform into novel solutions that our clients actually require. So bringing all this together, I think, really does give us an element of competitive advantage.
[00:12:02] And I think you asked also, how do they leverage AI and data science? Well, that's really across the board. It's almost easier to ask which parts don't leverage AI and data science. Well, you know, there aren't any. So really the whole value chain from onboarding of data, mapping of data, quant research, which fundamentally is financial data science, I suppose, portfolio construction, best execution, risk analysis.
[00:12:31] It's absolutely across the board. It's ingrained into everything that we do. And of course, AI gets all the headlines. It gets all the attention. But one of the things we don't talk about enough is that data is the lifeblood of AI. And indeed, data is also at the very core of asset management. So with both those things in mind, tell me more about Man Group's latest data research projects and maybe how they're helping shape the future of trading initiatives too.
[00:12:59] Yeah, I think you're absolutely right to view data like that. It's absolutely core. And I kind of view every team within the firm, and this is not just the investments teams, but every single team. But what are we doing? We're bringing together some type of professional expertise, hopefully like good tech and good tools. But data is going to fuel that, right?
[00:13:23] Data inputs, actionable insights, data outputs that can go on to inform other teams or other parts of the investment process. So very much data, yeah, first order concern within the firm. And loads going on. We could be happily chatting, I think, all day if we were to really go into everything that we're doing in any amount of debt. But maybe a few highlights for you and listeners that will give a bit of a taste.
[00:13:51] So first taster. For some time now, we've very much viewed a lot of the way we compose our trading strategies as computation graphs, many thousands of individual little white nodes which perform some mathematical calculation, perhaps each with their own data inputs, each with their own data outputs. So we've been doubling down on this approach with a project called Pondor.
[00:14:17] That brings what has evolved over the past decade, lots of different ways of building trading strategies that ultimately deal with data into one real cross-asset class graph-based big quant research and trading framework. And that's also doing this from the ground up. And it's very much a once in a decade thing, doing something like this.
[00:14:43] We've also tried to fundamentally engineer from the ground up better ways of approaching and dealing with data. So data provenance is now very much a first-class thing in the platform, understanding where data has come from, understanding all of the transformations that have occurred to it. So we can backtrack and explain everything that has happened along the value chain. So that's proving to be really powerful.
[00:15:11] And then we get some lovely trading system and data flow visualizations out of this that look like a big complex, like I'm a physicist by background as well. I like cosmology. So big star clusters of computation nodes and edges connecting everything together. So that has been a whole lot of fun. That's kind of one thing. Other things like low latency data, level three data, markets by order data.
[00:15:38] And looking at how we might better improve our execution capability. And that's an interesting data challenge because the data quickly becomes very, very voluminous, like multiple petabytes of data to deal with there. And looking at building our code capability.
[00:16:01] Again, better execution capability gives us the ability to monetize signals that perhaps we wouldn't have been able to before because the alpha decay in them is perhaps quite quick. So a lot of focus there, level three tick data and color. I think another thing I'd say, in a way I often think about data in this type of industry is in the form of a letter T.
[00:16:30] If we visualize a capital letter T and for traditional quant, that's often the bar of the T, right? It's the, we want very broad sets of data that cover lots of instruments and we're looking for broadly applicable signals and insights. But for dedicated discretionary portfolio managers, what might they want? They might want very deep insights in a particular company or in a particular little industry sector.
[00:16:59] And that might cause us to want to build like web scraping like tech to meet the needs of those PMs. But another thing we're trying to do, if that's the, I guess, a historical type of model of how to look at things, let's fill in the gaps underneath the bar of the T as well. With sector specific, industry specific abilities for research and onboarding.
[00:17:22] So all of this, of course, this type of data research, this type of building, composing strategies in a way that requires more data or requires us to look at data differently or to map data in a different way. Of course, more demand on engineering, compute, infrastructure, onboarding, operations. So kind of back to the, what does tech allow us to do?
[00:17:48] Well, it allows us to innovate, but we've got to do so efficiently in order to operate at the scale we want. It's kind of back to exhibit A, the tech platform piece from a minute ago. And I must admit, after Googling you before you came on the podcast, I learned as well as being the CTO for Man Group, you're also running ArcticDB, which is a tech startup that will soon be signing your 10th paying customers, I believe. So what's the story here? There's got to be a story there, right?
[00:18:15] There is a story behind ArcticDB, yeah, and smiling. So look, we're open source first a lot of ways. We talked about that a moment ago. And we're very much a, we're an investment company. Our job is to deliver returns to our clients. It's an old stand up in front of the team. I know it's fashionable for some investment leaders to say, oh, we're a tech company now.
[00:18:45] And I'm thinking, no, you're not. You're just saying that because it's fashionable. I know people who work for you and you're not a tech company. So Man Group, we are an asset manager. Our job is to build returns for clients. Our job is not to build like tech for tech's sake. So why on earth have we built a database and why are we trying to sell that? It's a good question. I said we're open source first. It's not something we set out deliberately to do.
[00:19:11] But when there's a gap, where there's a need for some technology that we don't think is adequately served by open source, then, you know, damn right we'll build it. We've got the technical bench here to do pretty much anything that we put our minds to. So fundamentally, Arctic DB, we call it a data frame database or a data science database.
[00:19:34] And if we think about the value chain in the world of investing from data in the outside world that we can bring on board and hopefully give us some insight into how the world behaves and how the prices of assets might be affected. Through all of the factors, the signals, portfolios, the whole value chain, regardless of how that data starts. It could be text. It could be imagery, perhaps in some cases.
[00:20:03] And it can be numbers. Of course, it's often numbers. But very quickly, we're into this world of data frames. Vast matrices of data and our models can rely on thousands or tens of thousands of these data frames. And existing tech to handle data frames at scale just wasn't there. And when I say it's scale, I, of course, we're all used to dealing with perhaps like tick data files, which are billions of rows long.
[00:20:30] But look for tech which can deal with matrices that are data frames that are hundreds of thousands of columns wide and to deal with things like that performantly. And there are real use cases that if you want to have a data frame which has got every corporate bond in existence and a column for each of those bonds, it can get pretty big pretty quickly.
[00:20:53] So ArcticDB was built to handle data at that type of scale and to get it from storage to distributed compute into Python as quickly as we possibly could. And I don't think there's anything out there which is faster. It will serve up a billion, not a million, a billion rows a second into Python. And that's why we built it. But in terms of why should we sell it?
[00:21:20] Because we could just choose to keep this proprietary and not open it up at all. Again, my view of tech is always advancing. If there's a need that we have, that Man Group has, there's likely to be a need that another firm has or another one firm today, 10 firms tomorrow, 100 firms in three years' time. And where needs arise, of course, things fill the gap.
[00:21:44] And if Arctic isn't the thing that fills the gap, then something else will fill the gap. And this proprietary competitive advantage today that we keep secret would just become something that's like legacy and deprecated tomorrow.
[00:22:01] So the hope very much is here that Arctic really becomes the de facto standard for dealing with data frames and doing data science at scale in this way. And so making it open source and building out the commercial offering on top just felt like the right thing to do. And listening to you today, it's clear that AI is already transforming the financial industry, particularly in asset management.
[00:22:30] But you also strike me as someone that's always got their eye on the future and what's ahead. So where do you see the biggest opportunities for AI and hedge fund investing over the next three to five years? And I understand it. The current climate, three to five years, is currently a lifetime away, isn't it? It's the speed of change. I think we talk about transforming the industry. I'd probably temper that a little bit. What's the reality?
[00:22:57] Well, certainly there's loads of hype. And I probably separate the hype into two threads of hype. There's the big tech thread of hype, which is perhaps at some sort of zenith now. That's notwithstanding deep seek disruptions that have happened relatively recently. And then I think there's the real world. What does it mean for firms like us? And what's the hype cycle like that?
[00:23:26] And I think for many people I talk to, and I'm pretty well connected, at least in London with industry peers, we're probably some way into the trough of disillusionment when it comes to AI. It's not getting quite the same level of airtime around the table, the dinner table that it had 12 months ago.
[00:23:49] And I think possibly some of the expectation that was set by the very impressive features of things like ChatGPT, that resulted in perhaps a over-inflected expectation that there'd soon be some kind of killer app in the workplace. And I think fair to say nobody's really seen or built any super killer finance app yet. And if they have, they're probably keeping it very secret.
[00:24:18] So I think that's perhaps the broader industry backdrop at the moment. But I do remain a massive proponent. I do remain very, very positive about this. I see an awful lot of incremental value add in the first instance. So pretty much every role, every job function, the ability to augment what we do, how we think to reduce toil.
[00:24:45] I mentioned a few minutes ago that ManGPT is in use by the majority of the firm here on a regular basis. And so very much, if we look at, say, the elite athlete model of how they improve their game, well, they're looking at all of the different constituent parts, they're refining things, they're making every little bit better. Now, arguably, that's not as exciting as some grand revolutionary change, but it's very real.
[00:25:12] It's very important nevertheless. I think it's important not to overlook that. But I guess the revolutionary stuff, the exciting stuff, one opportunity I see for disruptive change is in the creative space, the quant space, the research space. I don't accept a premise that, hey, this is a creative process, nothing to see here. Just build me some more tech tools and let me get on with it.
[00:25:40] But fundamentally, if we look at what does tech do, how does it add value? Well, I'd argue that we add value by a repeatable process of understanding workflows quite intimately and then automating those workflows. So quant and technology, that's allowed us to systematize investment processes, right? So the next level to that might be to think about how we can systematize, how we build the systematic strategy.
[00:26:10] So if we start to break down that creative process, that quant research process into constituent parts and start to think about how we can build some of the new advanced reasoning models to bear there, I think we may be able to do just that. So I'd say perhaps any slightly tong and she, quant, watch out. Take what we're coming for you next. And I think that's quite exciting.
[00:26:32] And we're doing a bunch of work there in the firm, by the way, with lots of internal projects, Alpha Assistant and Alpha GPT. And one of the components of Alpha GPT is Idea GPT designed to potentially disrupt some of those creative elements of the investment process. So I'm excited to see where that goes. Love that.
[00:26:58] And of course, as AI becomes more prominent in trading strategies, on the flip side of that, I've got to ask, what kind of challenges do hedge funds face when it comes to balancing automation with human expertise? Again, it feels like quite a big balancing act. It does. Yeah, it's a good question. There are certainly some new challenges here or perhaps some new takes on existing challenges. And perhaps some of the existing challenges just got a bit bigger.
[00:27:27] And for quite a long time there, particularly in the quant hedge fund space, it's been described as black box. But at least in this firm, that's never really been true. And we have all sought to understand every stage of the decision-making process and for every single computation throughout the diagnostics. So we can go back and explain why things happened how they did.
[00:27:52] I think AI does tend to take that challenge of explainability to a new level when things are much more complex or can be much more complex than they were. And if we're relying on external models, then often they're closed both in terms of source code and the training, the weights that are part of the model.
[00:28:13] Interesting to see some of the advances, I think, here, how we can now start to ask models to explain their reasoning step by step by step. I think that's important. Because fundamentally, in the event of an error, I don't think in this industry it's acceptable to say, oh, we don't have any rationale for that trade or that position. But the model told us to do it. I don't think that's a tenable one to have.
[00:28:41] It's not one I'd be comfortable saying to our CEO or one of our CEOs, hey, the model just told us to do this. And I guess that leads to a view that Gen AI, at least right now, and we're talking about Gen AI specifically, a lot of the use cases are either very much like copilot, human augmentation, not necessarily autopilot, though.
[00:29:07] While we are turning the autopilot on, I think very much we need to be able to look to explain what we're... Other side of this, of course, is data. We've talked about data quite a bit. Data quality is also important. If we're putting data into processes and automating things and getting insights out, then let's know what we're putting in.
[00:29:30] But again, dealing with data, dealing with dirty data, and the world of data is a... It always has been a horribly dirty world to deal with. Mistakes, inaccuracies, badly formed things, missing data, restated data. I mean, it's something we're used to dealing with and have been used to dealing with in the industry for a long time.
[00:29:57] So maybe the problem gets bigger, challenge gets bigger, but fundamentally, it's an area that I think we're... We are well-placed to deal with. And on a more personal level, given your background from hobbyist coder to CTO, I've got to ask, how has your experience influenced Man Group's technology strategy, especially in areas like front office technology and data-driven decision-making? Yeah, you mentioned data-driven decision-making.
[00:30:26] So as well as a hobbyist coder, I'm a physicist by training. There are a lot of physicists actually in the tech team at Man, and we joke between ourselves that physicists act, physicists actually make the best tech folks. I think the computer scientists might disagree, and I think it's good to have some diversity. But in all seriousness, that background in physics has certainly led to very much a data-driven approach to decision-making, decisions that I make.
[00:30:55] And I think augmenting that with the background as a hobbyist coder, and in my teenage years, I was hacking assembler, I was hacking machine code. So that's been a really important building block that's stood in good stead, I think. So definitely there are people much better equipped nowadays to deal with writing production trading code than I am. I'm not sure you want to let me super close to the code base of the system.
[00:31:23] I do know what good looks like, and I know what's going on right down to the metal. I think having that knowledge and understanding is useful. It does stand on in good stead for any type of broad CTO-type position. And if we do dare to look ahead, how do you envision technology and AI continuing to evolve within asset management? And what should firms be doing right now to stay competitive in this increasingly AI-driven world?
[00:31:52] Because it's only going to keep going faster and faster, isn't it? I think it is. I think you've got to embrace it to start with. Don't just stand on or rabid in headlights that are scared of it. I think it acknowledged the fact that even if you are in some part of that trough of disillusionment, because you're not seeing the killer app today, all change like this.
[00:32:16] If we look through history, it's always taken a long time to really bear the fruits in industry. So embrace it for sure. I think, again, back to the perennial buy versus build debate, the resources are always finite. There's always more demand for tech than there are people to do it, it seems.
[00:32:41] I think think long and hard about what you buy, what you build, what we build today might be commodity tomorrow. We've seen that with the Gen AI space moving so quickly. Quickly, I've had many conversations with the head of machine learning tech in the firm and he'd be thinking of something on a Friday. And over the weekend, something's been released which has upended the thing he was building or the thing he was thinking of building.
[00:33:09] So I think another bit of advice would be to frame its look to open source. This combination of many, many cohesive building blocks combined with smart engineers and your own USP is a really powerful model. I think the final thing I'd say, another way I look at tech is, and this is invariably one of my interview questions when I'm talking to somebody about joining the team,
[00:33:37] is to tech game of two halves. There's building the thing right, building the thing in a well-engineered way. But we've also got to build the right thing in the first place. And building the right thing involves asking the right questions or putting the right prompts into the LLM. And I think it's increasingly imperative that engineers actually understand finance,
[00:34:04] understand the problem, know the why of the thing that they're doing. Because coding copilots, augmentation, that's becoming increasingly impressive. And just being able to code something up and not necessarily understand why that's the case, that's increasingly, I think, a commodity type of skill set. Yeah, I completely agree with you.
[00:34:32] And I cannot thank you enough for taking the time to sit down and share your insights with me today. But before I do let you go, I want to have a little fun with you and ask you, Steve, a few more final gifts for everyone listening. And that is a book for our Amazon wishlist and a song for our Spotify playlist. I don't mind what you add there. Guilty pleasures are allowed. But what would you like to add and why? So the book I would add, and I did scan through your list.
[00:35:01] I don't think it's there already. Forgive me if it is. The book is called Range by David Epstein. And that's got a subtitle of why generalists triumph in a specialized world. And really, if we're to sum up this book, for me, its thesis is if you want to solve really hard, knotty problems, the best way to do that is breadth wins out over depth. Cross-fertilization between domains is vitally important.
[00:35:29] So maybe in some ways it's a bit of a validation of how I view my job, this big multidimensional optimization problem, and I guess supports the collaborative culture that we have and seems to work well within the firm. So yeah, Range by David Epstein would be the book. In terms of I'm a huge Lana Del Rey fan. In fact, probably 80% of my entire listening is listening to Lana. But I'll pick out one song, and that's Old Money.
[00:35:58] It conjures up some really lovely West Coast U.S. images and history. So I think that's the song I choose. Oh, fantastic. I will get that book added to our Amazon wishlist, and indeed the song, great song too, added to the Spotify playlist. And for anyone listening wanting to find out more information about the many different plates that you spin, the hats that you wear, and your work at Man Group, where would you like to send everyone listening?
[00:36:26] I, we, firm, big proponents of open source. We've been doing open source a long, long time there, giving back when it wasn't really fashionable to do so. Go look at GitHub. Go look at Man Group on GitHub. Go look at some of our code. Go download ArcticDB. Take it for a spin. I think that's a good way of getting to grips with all the tech teams doing. Look at the code that they're writing and give it a whirl. Fantastic advice. I will leave links to everything so people can find you nice and easily.
[00:36:56] And well, I cannot thank you enough really for shedding light on the technical trading initiatives underway at Man Group today, and also learning a little bit more about looking under the hood, how AI is being used across the firm, how it's impacting hedge fund investing more broadly, and what we expect in the future, not to mention a great book and a song too. I think we covered everything. So thank you so much for talking with me today. Thank you, and my pleasure.
[00:37:21] So from leveraging AI-powered trading models to developing open source data solutions like ArcticDB, Gary has provided fascinating insights today into how Man Group is using technology to stay ahead in a highly competitive industry. And one of the things that stood out to me is their commitment to balancing automation with human expertise, particularly in areas where AI explainability is critical.
[00:37:48] I think this highlights the complex yet promising future of AI in this industry. And as AI moves beyond operational efficiencies and begins to influence the creative side of quantitative research, there are new questions that are beginning to arise. How far can AI push the boundaries of financial innovation? Can we really trust AI models to make decisions without fully understanding the reasoning behind them?
[00:38:15] And what role will human judgment play in this increasingly automated investment world? You've heard from me. You've heard from Gary today. Now it's time for me to put the microphone in front of you. I'd love to hear your thoughts on this. Is AI ready to take on the creative challenges of asset management? Or does human expertise remain irreplaceable? Or most likely, is it a combination of all those things? Email me, techblogwriteroutlook.com
[00:38:44] Instagram, x, just at Neil C. Hughes And my website, if you want to work with me, is techblogwriter.co.uk But that is it for today. It's quitting time for me. Big thank you to Gary for sitting down with me. An even bigger thank you to each and every one of you for tuning into this podcast every day. But have no fear. I'm not going to be away that long. I'll be back again bright and early tomorrow, lingering in your podcast feed, just daring you to hit play one more time.
[00:39:13] I will speak with you all then. Bye for now.