New Research by The Alan Turing Institute Highlights Potential of Large Language Models in Finance

New Research by The Alan Turing Institute Highlights Potential of Large Language Models in Finance

(IN BRIEF) The Alan Turing Institute has released a groundbreaking report suggesting that Large Language Models (LLMs) could revolutionize efficiency and safety within the finance sector. The research explores how LLMs can detect fraud, generate financial insights, and automate customer service tasks, showcasing their potential to enhance services across various sectors, including banking, insurance, and financial planning. Conducted through a literature survey and a workshop involving professionals from major banks, regulators, insurers, and government agencies, the study reveals that LLMs are already being utilized to streamline internal processes, improve decision-making, and enhance productivity. While participants foresee LLM integration into investment banking and venture capital strategy development within two years, they also acknowledge the technology’s associated risks, particularly concerning regulatory compliance and explainability. The report recommends collaborative efforts among industry stakeholders to address safety concerns and advocates for further exploration of open-source models while prioritizing security and privacy considerations. Professors Carsten Maple and Lukasz Szpruch from The Alan Turing Institute underscore the importance of industry-academic collaboration in assessing the potential and challenges of LLM implementation in the highly regulated financial sector.

(PRESS RELEASE) LONDON, 27-Mar-2024 — /EuropaWire/ — Large Language Models (LLMs) have the potential to improve efficiency and safety in the finance sector by detecting fraud, generating financial insights and automating customer service, according to new research by The Alan Turing Institute published today (Wednesday 27 March).

Because LLMs have an ability to analyse large amounts of data quickly and generate coherent text, there is growing understanding of the potential to improve services across a range of sectors including healthcare, law, education and in financial services including banking, insurance and financial planning.

This report, which is the first to explore the adoption of LLMs across the finance ecosystem, shows that people working in this area have already begun to use LLMs to support a variety of internal processes, such as the review of regulations, and are assessing its potential for supporting external activity like the delivery of advisory and trading services.

Alongside a literature survey, researchers held a workshop of 43 professionals from major high street and investment banks, regulators, insurers, payment service providers, government and legal professions.

The majority of workshop participants (52%) are already using these models to enhance performance in information-orientated tasks, from the management of meeting notes to cyber security and compliance insight, while 29% use them to boost critical thinking skills, and another 16% employ them to break down complex tasks.

The sector is also already establishing systems to enhance productivity through rapid analysis of large amount of text to simplify decision making processes, risk profiling and to improve investment research and back-office operations.

When asked about the future of LLMs in the finance sector, participants felt that LLMs would be integrated into services like investment banking and venture capital strategy development within two years.

They also thought it likely that LLMs would be integrated to improve interactions between people and machines, for example dictation and embedded AI assistants could reduce the complexity of knowledge intensive tasks such as the review of regulations.

But participants also acknowledged that the technology poses risks which will limit its usage. Financial institutions are subject to extensive regulatory standards and obligations which limits their ability to use AI systems that they cannot explain and do not generate output predictably, consistently or without risk of error.

Based on their findings, the authors recommend that financial services professionals, regulators and policy makers collaborate across the sector to share and develop knowledge about implementing and using LLMs, particularly related to safety concerns. They also suggest that the growing interest in open-source models should be explored and could be used and maintained effectively, but that mitigating security and privacy concerns would be a high priority.

Professor Carsten Maple, lead author and Turing Fellow at The Alan Turing Institute, said: “Banks and other financial institutions have always been quick to adopt new technologies to make their operations more efficient and the emergence of LLMs is no different. By bringing together experts across the finance ecosystem, we have managed to create a common understanding of the use cases, risks, value and timeline for implementation of these technologies at scale.”

Professor Lukasz Szpruch, Programme Director for Finance and Economics at The Alan Turing Institute, said: “It’s really positive that the financial sector is benefiting from the emergence of large language models and their implementation into this highly regulated sector has the potential to provide best practices for other sectors. This study demonstrates the benefit of research institutes and industry working together to assess the vast opportunities as well as the practical and ethical challenges of new technologies to ensure they are implemented safely.”

Media Contact:

Fiona Dennehy
Senior Media Relations Manager
fdennehy@turing.ac.uk
(0)20 3862 3598

SOURCE: Alan Turing Institute