Nova: Picture this: you're a portfolio manager at a major investment firm. Every morning you wake up to a tsunami of data — earnings calls, regulatory filings, breaking news, social media chatter, market prices updating by the millisecond. Somewhere in all that noise, there are signals. The question is, can you find them before your competitors do?

Nova: : That sounds overwhelming. I mean, even with a team of analysts, how could anyone process all that?

Nova: That's exactly the problem Joseph Simonian tackles in his new book, AI in Asset Management: Tools, Applications, and Frontiers. Published by CFA Institute Research Foundation in 2025, it's essentially a field guide for investment professionals trying to navigate what Simonian calls the "data fog."

Nova: : The data fog — I love that phrase. So this is a book about using artificial intelligence to cut through it?

Nova: Exactly. But here's what makes it different from every other AI book out there. It's not about the hype. It's about what actually works when you sit down at your desk on Monday morning. Simonian brought together thirteen global experts — people from Fidelity, Citi, PGIM, JPMorgan, NYU — and they each wrote a chapter on the AI technique they actually use. It's the shift from theory to practice.

Nova: : So this isn't some academic exercise. These are people in the trenches.

Nova: Right. And Simonian himself is fascinating. He's got a PhD from UC Santa Barbara, he's spent 20 years at firms like Fidelity, Natixis, and J. P. Morgan, and he now runs his own AI-driven investment firm, Autonomous Investment Technologies. He's also the guy who literally wrote the CFA Institute's report on "AI washing" — you know, when firms claim to use AI but are really just running Excel macros.

Nova: : AI washing! That's a real thing?

Nova: Oh, it's huge. Simonian has called it out repeatedly. The barriers to real AI adoption in investing are genuinely high — financial data is messy, sparse, and hard to predict. But that's exactly why this book matters. Let's dive in and see what separates real AI from the noise.

Nova: So here's the big idea that runs through the entire book. Traditional quantitative finance is built on prediction — estimating returns, forecasting volatility, calculating default probabilities. But knowing what might happen and knowing what to do about it are two completely different things.

Nova: : That sounds obvious, but I've never heard it framed that way before. Most quant work really is just about making better forecasts.

Nova: Exactly. And Simonian's book argues that's insufficient. One of the most compelling chapters is by Igor Halperin from Fidelity, Petter Kolm from NYU, and Gordon Ritter, who was named Buy-Side Quant of the Year. They make the case for reinforcement learning — teaching algorithms to make sequential decisions in uncertain environments.

Nova: : Reinforcement learning — like the same technology that beat world champions at Go and chess?

Nova: Same family, but applied very differently. In finance, decisions compound. A trade you make today changes the prices you'll face tomorrow. A hedge you put on now might only prove its worth months later. Traditional models can't handle those feedback loops. But reinforcement learning was literally designed for this.

Nova: : So instead of saying "this stock will go up 5%," the AI learns "here's how much to buy right now to maximize long-term returns while managing risk."

Nova: That's it. And here's the really clever part. They also introduce inverse reinforcement learning. Instead of teaching an AI what to do, you observe what skilled human traders actually do, and the AI reverse-engineers their hidden objectives. It learns intent from behavior.

Nova: : Wait, so the AI watches a star portfolio manager and figures out what they're really optimizing for, even if the manager can't articulate it?

Nova: Precisely. Halperin, Kolm, and Ritter describe using this for everything from trade execution to option hedging to market surveillance. They even suggest regulators could use it to detect market manipulation by inferring trader intent from order flow.

Nova: : That's both brilliant and slightly unsettling. So the book is advocating for a fundamental shift in how finance thinks about AI — not as a prediction machine but as a decision engine.

Nova: Let's walk through what's actually in this book. Simonian organized it like a toolkit. Ten chapters, each one focused on a different AI technique with proven investment applications.

Nova: : A greatest hits of machine learning for finance.

Nova: That's a good way to put it. Chapter one is Simonian's own overview of unsupervised learning. This is AI that finds hidden patterns in data without being told what to look for. In finance, that's huge because markets are opaque — you often don't know what the "right answer" is.

Nova: : Give me a concrete example.

Nova: Sure. He discusses using something called spectral clustering to build a regime-based trading model. Basically, the algorithm discovers that markets are operating in different "regimes" — risk-on versus risk-off, high volatility versus low volatility — and then you trade accordingly. Simonian and a co-author actually published a paper showing this framework produces economically intuitive signals for growth, inflation, and leverage.

Nova: : So the AI figures out what kind of market we're in without anyone labeling it.

Nova: Exactly. Then chapter two, by Gueorgui Konstantinov and Agathe Sadeghi, extends this into network theory — treating financial markets as a complex web of interconnected assets rather than isolated securities. This is how you start to understand systemic risk.

Nova: : What about the hot topic — deep learning?

Nova: Chapter five, co-authored by Simonian and Paul Bilokon from Imperial College London. They show how neural networks can price complex derivatives in milliseconds while providing real-time risk estimates. One specific application: pricing options with "stable Greeks" — those sensitivity measures that tell traders how their positions will react to market moves. Traditional models struggle with this; deep learning handles it gracefully.

Nova: : And natural language processing — that must be a huge chapter given everything happening with ChatGPT and large language models.

Nova: Chapter seven, by Francesco Fabozzi from Yale. It's fascinating. He traces NLP's evolution from counting positive and negative words in earnings calls — imagine a simple dictionary lookup — all the way to today's large language models that can summarize entire 10-K filings, extract ESG risks, and monitor compliance in real time.

Nova: : But Fabozzi also warns about hallucinations, right? LLMs can produce things that sound authoritative but are completely wrong.

Nova: Absolutely. He emphasizes that governance, evaluation frameworks, and audit trails are just as important as the models themselves. The book doesn't shy away from AI's limitations — it confronts them head-on. That's actually one of its greatest strengths.

Nova: Let's talk about chapter ten — Ethical AI in Finance, by Anna Martirosyan from EY Parthenon. This isn't just a "nice to have" afterthought tacked onto the end. It's central to the book's thesis.

Nova: : This feels particularly important given everything we've discussed. You can't just unleash black-box algorithms on people's retirement savings.

Nova: Exactly. And CFA Institute has been leading this conversation. The book reinforces their position that AI must be deployed with transparency, accountability, and professional integrity. Martirosyan's chapter addresses questions like: How do you ensure an AI model isn't embedding bias? What happens when an algorithm makes a catastrophic error — who's responsible?

Nova: : These aren't hypothetical questions anymore.

Nova: Not at all. And Simonian has been particularly vocal about "AI washing" — the practice of firms claiming to use sophisticated AI when they're really doing basic statistical analysis. He published a separate CFA Institute report on this, laying out specific questions investors should ask to evaluate whether a manager's AI claims are genuine.

Nova: : What kinds of questions?

Nova: Things like: What specific machine learning techniques are you using? How do you handle overfitting? Can you explain how your model arrives at decisions? What's your data pipeline look like? If a manager can't answer those, you're probably looking at AI washing.

Nova: : So the book is essentially equipping professionals to be intelligent consumers of AI, not just users.

Nova: That's exactly right. Mona Naqvi, Managing Director at CFA Institute, wrote in the foreword that "the transformative power of this technology must be unlocked in ways that strengthen, not supplant, human judgment, trust, and fiduciary responsibility." The book's consistent message is that AI should augment human expertise, not replace it.

Nova: : There's something refreshing about a technology book that doesn't promise the moon. It's saying: here's what AI can realistically do, here's where it fails, and here's how to deploy it responsibly.

Nova: The book doesn't just cover what's working today. It also looks at the frontier. Chapter nine, by Oswaldo Zapata, is on quantum computing for finance.

Nova: : Quantum computing in a practitioner book? That seems wildly futuristic.

Nova: It is, but Zapata makes the case that certain financial problems — like portfolio optimization with thousands of assets and complex constraints — are computationally intractable for classical computers but potentially solvable with quantum approaches. He's not saying it's ready for prime time. He's saying: here's what's coming, start paying attention.

Nova: : And chapter eight covers machine learning in commodity futures, right?

Nova: Yes, by Tony Guida, a quant portfolio manager at Atonra. He shows how ML can bridge data, theory, and return predictability in commodity markets. These are notoriously difficult markets to model because they're driven by everything from weather patterns to geopolitical events to supply chain disruptions. Machine learning excels at finding structure in that kind of chaos.

Nova: : What about ensemble learning? I've heard that term but never quite grasped it.

Nova: Chapter four, by Alireza Yazdani from Citi. The idea is beautifully simple: instead of betting on one model, you combine many models and let them vote. It's like getting a second opinion, and a third, and a fiftieth. In investment contexts, ensemble methods are particularly powerful because no single model captures all market dynamics. When one model misses something, another catches it.

Nova: : So the book really does span from the fundamentals to the bleeding edge. What's Simonian's overall vision for where this is going?

Nova: In the preface, he's quite humble about it. He says the book is "not the conclusion of the financial data science story but rather a chapter written during a pivotal moment." He compares the contributors to jazz musicians — each has their own style, their own voice, but they're all playing within the same rhythm. The unifying theme is how AI deepens our understanding of markets.

Nova: : A jazz ensemble of quants. I love that image. It suggests creativity and improvisation, not just cold calculation.

Nova: Exactly. And that's the heart of the book. AI isn't about replacing human judgment with algorithms. It's about expanding what human judgment can do. The most effective investors of the next decade will be those who combine deep domain expertise with data-driven adaptability.

Nova: So let's bring this together. Joseph Simonian's AI in Asset Management is really three books in one. First, it's a technical reference — covering everything from unsupervised learning and support vector machines to deep learning and reinforcement learning, each written by someone who deploys these tools professionally.

Nova: : Second, it's a strategic playbook. It helps leaders figure out where AI actually adds value versus where it's just window dressing. Simonian's work on AI washing is a perfect complement — teaching professionals how to separate signal from noise in AI claims, not just in market data.

Nova: And third, it's an ethical compass. The book insists that governance, transparency, and human oversight aren't optional extras. They're prerequisites for any serious AI deployment in finance.

Nova: : If I'm a portfolio manager or analyst listening to this, what's my one big takeaway?

Nova: Start with a specific problem, not with a cool algorithm. The book's clearest message is that successful AI adoption begins with identifying use cases where machine learning adds demonstrable value beyond traditional quantitative methods. Then you build the infrastructure, the governance, and the talent around that problem. Don't start with the technology and go looking for a problem to solve.

Nova: : And for someone who's not a quant — maybe a compliance officer or a product strategist?

Nova: You need to understand enough to ask the right questions. What's the data pipeline? How is the model validated? What are the governance protocols? Where could it fail? The book equips you to have those conversations intelligently, even if you're not coding the models yourself.

Nova: : One thing that struck me throughout our conversation: this isn't a book about AI replacing humans. It's about humans becoming better at what they do.

Nova: That's it exactly. As Simonian writes, AI's greatest promise lies not in replacing human judgment but in enhancing it. The goal is to cut through the data fog, to harness AI as a guide through complexity rather than a source of it. In an industry drowning in information, that's not just valuable — it's essential.

Nova: : This is Aibrary. Congratulations on your growth!