AI for Financial Services

This workshop was organized as an opportunity to foster collaboration between experts in the field of financial services and AI, from both academia and industry. Three important topics were discussed:

This was defined as exploration of AI methods and tools beyond large language models (LLMs). The example of the SORA algorithm was presented. It was highlighted new specialist high-fidelity, robust AI-based simulators for financial services will need to be able to deal with complex dynamical systems and large multi-modal datasets.

The discussion focused on how to extract the most value out of unstructured data, enhance multi-modal forecasting and the ability for AI to analyse publicly available textual information for Environmental, Scoial and Governance (ESG) ratings. Aside from accruacy and reliability, the need to make LLMs more autonomous (less dependent on prompt) and fliexible was highlighted. A number of applications were identified, such as automated customer services with personalised AI assistants and human experts in the loop. It was stressed that to successfully integrate LLMS into financial services, language and jargon bias as well as lag due to limited hardware resources will have to be addressed.

The nature of the regulatory problem was noted as a global issue requiring international collaboration. The need for the development of explainable AI was noted as crucial, but also to ensure the traceability of the original datasets used. The question of liability was discussed and it was highlighted that full implementation of AI cannot be achieved by the financial service industry, even if perfect models exist, until regulation issues are resolved.