TECH TUESDAY is a weekly content series covering all aspects of capital markets technology. TECH TUESDAY is produced in collaboration with Nasdaq.
Generative artificial intelligence has all the buzz these days, but when it comes to market surveillance, how does the emerging technology translate from conceptual and abstract, to concrete and specific?
That question was explored in a recent Nasdaq webinar, Practical Applications of Gen AI in Surveillance. Tony Sio, Head of Regulatory Strategy and Innovation, Nasdaq, and Ruben Falk, Generative AI and Machine Learning Lead, Financial Services, Amazon Web Services discussed what gen AI is, practical applications in surveillance, and how AI can improve and drive efficiencies in a mission-critical function for brokers and market operators.
Falk set the stage by offering a standard definition of gen AI: a model that trains on unstructured data and other media, and is able to learn from that data in such a way that it can produce new artifacts. The bang for the buck is in the sheer volume of data that can be tapped; a large language model (LLM) has the knowledge of a human who’s been reading “everything under the sun” for 20,000 years, Falk said.
“But the limitation is that even though these models possess a lot of knowledge and have some reasoning capability, they’re ultimately probabilistic in nature and sometimes their answers may resemble the correct answer but actually be incorrect. This is also known as hallucinations,” Falk said. “This is an area that we’re focusing a lot on with Nasdaq and with our other customers, namely how to guard against hallucinations and optimize for accuracy.”
It was noted that some use cases for traditional machine learning are morphing into use cases for gen AI.
“For instance, algorithms that detect a potential fraudulent activity or compliance breach, those types of algorithms are migrating from rules-based systems to traditional machine learning systems, but not really generative AI systems,” Falk said. “The idea of making a decision or a prediction is not really what generative AI is all about. However, generative AI is being used to produce inputs for the traditional machine learning algorithms.”
Falk and Sio agreed that surveillance is fertile ground for gen AI, as it greatly expands the universe of potential inputs for market abuse detection. Applying generative AI to surveillance can help improve both effectiveness and efficiency when it comes to navigating growing and often unstructured data sets, evolving behaviors, and complex trading activities across markets and assets classes, among other things.
But the technology also offers broader, enterprise-level utility.
Nasdaq frames gen AI use cases in two different ways. “One is how we can improve our products – our surveillance tools, our compliance tools, our trading with AI functionality,” Sio said. “And then we also look at it in terms of the business – how can we improve the Nasdaq business itself? How are we using AI internally?”
Nasdaq is deploying gen AI internally for coding companions and content creation, as well as discovery and workflow automation, Sio said. There’s also an initiative around AI education for employees.
“We’re actually doing a fairly significant upskilling process right now where we are requesting almost all Nasdaq staff to perform training this year on AI,” Sio said. “Hopefully a lot of people will pick generative AI as part of that training.”
Creating tomorrow’s markets today. Find out more about Nasdaq’s offerings to drive your business forward here.
