Can AI Agents Make Financial Markets Less Stable?

Knowledge at Wharton

Ask about Knowledge at Wharton

Episode Description

ABOUT THE EPISODE AI could make financial decisions faster and more efficient. But when multiple AI agents learn, trade, and interact at once, they may also create new risks for competition and market stability.

Winston Dou, Faculty Director of the Wharton AI in Finance Lab, explains how AI is transforming fraud detection, credit assessment, trading, and investment research. He discusses why regulators and financial institutions need to evaluate AI systems not only on individual performance, but also on how they behave together in the market.

ABOUT THE PODCAST Every day, business scholars answer pressing questions in their research - but what do their insights mean for you? In this podcast, Wharton faculty dive into what inspired their studies and how their findings resonate with the world today. Learn how research insights translate into knowledge you can use, with host Dan Loney.

----- The Wharton School of the University of Pennsylvania is committed to sharing its intellectual capital through Knowledge at Wharton, the school’s online business analysis journal.

Launched in 1999, Knowledge at Wharton offers free access to:

- Articles, podcasts, and videos highlighting Wharton faculty research and analysis of current business trends - Interviews with book authors from Wharton School Press - In-depth series of curated content like Wharton Executive Education’s Nano Tools for Leaders - A searchable database of more than 10,000 articles covering all aspects of business - A weekly newsletter that delivers Knowledge at Wharton insights directly to your inbox

Visit the Knowledge at Wharton homepage: https://whr.tn/3qUpSBE Subscribe to the Knowledge at Wharton YouTube channel: https://whr.tn/3DEAgAU

Read the full transcriptHide the full transcript
  1. AI-ready supports fraud detection, trading, customer service, and more importantly, as we have seen, the
  2. generative AI is extending how we get information and how we process information to support investment
  3. decisions. Welcome to the Ripple Effect, the podcast that takes you on a journey through the minds
  4. of work and faculty. I'm your host, Dan Loney, and in each episode, we'll be diving deep into the inspiration
  5. behind the groundbreaking research that Wharton professors have conducted and exploring how their findings resonate with
  6. the world today. We already know that many areas are being impacted by the influence of artificial intelligence.
  7. One that we speak of today is about the world of finance. The Wharton School recently started a program that
  8. looks at the connection of finance and AI and what this could mean in the future.
  9. Winston Doe is an Associate Professor of Finance here at the Wharton School. He is also Faculty Director of the AI
  10. in Finance Lab as part of Wharton's AI and Analytics Initiative. Winston, great to talk to you again.
  11. How are you, sir? Good. How are you, Dan? Thank you for the introduction. Thank you.
  12. Let's start out by talking about the importance of having this type of a project ongoing,
  13. especially when AI is in so many discussions these days. Yes. So I think now is definitely the right
  14. time for us to make progress, to think about how AI will impact the financial market,
  15. right? Because finance is a particularly important place to understand AI's impact because financial decisions affect nearly
  16. everyone. And the financial markets determine where capital goes, who gets credit, how risks are shared.
  17. As AI reshapes those decisions, the consequences extend across the economy. So research and education must keep pace with
  18. adoption. So our lab brings together researchers and students and the practitioners to understand these effects and
  19. turn that knowledge into better financial decisions. How much and where is AI already having
  20. an impact in the finance sector? I would say AI already had a very broad impact in the financial markets.
  21. AI already supports fraud detection, as example, and the credit assessment, trading, customer service, and more
  22. importantly, as we have seen, the generative AI is extending how we get information and how
  23. we process information and also how we execute the information to support investment decisions.
  24. And adoption varies across institutions and tasks, of course, but AI is already changing the daily
  25. work of finance. When you have a collaboration like this and you're looking at both AI and finance, then
  26. what do you think are some of the most important questions to ask right now? Yeah, there are actually, there are many interesting
  27. questions and important questions. I would highlight three. So one is, when and how does AI
  28. actually improve financial decisions? Second, how does it change competition and stability
  29. when multiple important institutions and many investors use AI technology simultaneously and when the AI agents
  30. interact in the markets? And the third one would be, who capture the benefits? Who suffer from this new technology?
  31. Consumers, investors, or a small number of powerful firms. So these questions connect technological progress to economic
  32. value. So like with many people right now, there are questions about how autonomous systems can oversee
  33. pieces of our day-to-day lives. How much do we want to have autonomous oversee the world of finance?
  34. Autonomy should depend on the task and the evidence of reliability and the potential risk of
  35. failure. So well-tested and bounded tasks can support greater autonomy. That's totally fine.
  36. And the more consequential decisions and in a more complex setting, the decisions need stronger oversight
  37. and intervention. So firms and the designated decision makers should remain comfortable for the systems they deploy and
  38. including technology supplied by someone else. And also I think it's important for us
  39. to think about the objectives and authority to act and how the autonomous systems interact.
  40. What kind of role will the AI agents play when they are incorporated into the finance
  41. system? Absolutely. So I think this AI agents can help a lot for us to improve the whole
  42. like information processing and execution decision-making process. So AI can help us extract information, build
  43. datasets and investigate questions that were previously too costly to study and to work on.
  44. There has to be though, from your perspective in the research that you're gonna be doing
  45. here with this new project, that there can be a great benefit from the connection of
  46. AI in terms of using and doing your own research, correct? Yes, exactly. So I think in general, in a fundamental
  47. sense, not only for research within the academic institutions, but also research in industry, I think
  48. AI lowers the cost of turning information into decisions and allows decision strategies to adapt through
  49. learning. But one point being emphasized in our own research is financial markets is very different from
  50. like the rigid rules in physics, in chemistry, right? It's quite responsive, quite endogenous.
  51. So financial markets will respond to those decisions and adoptions of AI technologies.
  52. And that feedback is central. So AI change, in other words, yes, AI will improve the information processing and the decision
  53. -making process. But at the same time, the feedback from the market to the adoption of these technologies
  54. will change the very environment that AI tries to understand. I've talked with some of your colleagues about
  55. the component of regulatory where AI and finance are concerned. When you think about the research that you're
  56. going to be doing, how does the component of regulatory fit in, do you think, to
  57. the dynamics of AI and finance working together in the future? The regulation is for sure necessary from my
  58. perspective and based on our research. Because, so this autonomous AI system, they try
  59. to, by design, they kind of like maximize the private benefit. But at the same time, an agent can
  60. perform well for its own owner. At the same time, it can also contribute to harmful market outcomes.
  61. So our research examines how learning agents can weaken competition or make market more fragile through
  62. certain form of collaboration or collusion or coordination. So we therefore need to evaluate their collective
  63. behavior as careful as their individual performance. So this is what regulation will be needed.
  64. How much do you expect this to be both a benefit to obviously the students there
  65. at Wharton, but also as you kind of alluded to, to the public markets and to companies within the finance sector as we move
  66. forward? Yeah, that's a great question. So for students, I think the goal is to combine the practical AI skills with a
  67. strong financial judgment. So we'll help the students to understand how the model works and how to test the
  68. model's outcomes or outputs and how to connect the model's outputs to decision-making.
  69. And for public markets, I think the one potential benefit is better information.
  70. And also on top of that, it can lower the cost of information extraction, processing, and
  71. it will help the public to make more productive investment decisions. Are you expecting the human component to be
  72. a constant part of this process moving forward? I think that's the discussion point is just
  73. how much AI will do and how much do we still need to rely on humans as a factor in the development of a
  74. lot of your research, but also the relationship between AI and finance moving forward.
  75. Well, yes, definitely. So human is always the crucial part because remember the AI technology is designed to serve
  76. human. Yeah. It's not designed to serve the AI itself, right? So because of this, the judgment of human,
  77. the monitor of human, the supervision of human will always be the crucial part. What do you hope then is the longer
  78. term benefit of having a project like this? So we hope the lab to produce lasting
  79. research and we can generate reliable integrated data, which is AI ready. And also we can produce the testing tools
  80. to help practitioners and the regulators to evaluate autonomous AI systems. Both benefits and risks.
  81. And at the same time, a long-term goal is to create a place where faculty, students and practitioners can communicate, collaborate and study
  82. how AI is changing finance and help shape its development. So basically we want to create a platform
  83. that connect efficiently, connects four areas. Research, AI finance infrastructure development, teaching and engagement
  84. with industry. Winston, great to have you with us today. Thank you very much for your time and
  85. all the best with the program. Thank you so much, Dan. Thank you. Winston Do, who is an associate professor of
  86. finance here at the Wharton School. And as we mentioned, he's the president and faculty director of the AI in Finance Lab.
  87. Thank you for listening to the Ripple Effect. We hope you found this episode informative and
  88. engaging. Don't forget to subscribe and leave us a review so that we can continue to bring
  89. you the best insight from the Wharton School.

Brief Summary

Winston Doe discusses AI's impact on finance, decision-making, and the need for regulation and human oversight.

Key points

  • The Ripple Effect Podcast Introduction. Join host Dan Loney as he explores the intersection of AI and finance with Wharton professors.
  • AI's Impact on Finance. Winston Doe discusses how AI is reshaping financial decisions and market dynamics.
  • The Role of Human Oversight. Winston Doe highlights the necessity of human judgment in AI-driven finance.

Tension

45/ 100Some tension

Related episodes