As part of the seminar Inquire Europe arranged the 3rd Asset Owner Lunch with the aim to provide a platform for our asset owner members to exchange knowledge and experiences. Chairwomen Marie Briere introduced the lunch seminar and was pleased to note that participation in the group continues to increase.
Nils Everling, Portfolio Manager at Swedish national buffer fund AP4, gave a presentation of use cases of LLMs for asset owners. He highlighted that their use through APIs can increase internal efficiency and productivity in a variety of tasks. He provided examples of automating workflows and collecting and pre-processing data. He underlined their ease of use and consequent low implementation thresholds. The discussion that followed covered, among other topics, the reliability of LLMs in a production environment.
Professor Marcos López de Prado OMC, ADIA Lab and Cornell University, followed with a presentation on ”Lessons for Financial Machine Learning from Causal Factor Analys”. Professor López de Prado began his presentation by pointing out that the use of mis-specified models in the asset allocation process had a high potential of causing inefficiency and investments losses, as they make it challenging to correctly attribute risk. He specifically mentioned so called ”collider” and ”confounder” variables, which are easily overlooked in conventional statistical procedures, leading to model misspecification. In the last three decades, statisticians have developed numerous computational methods and algorithms for the discovery of causal relations, represented as directed acyclic graphs (DAGs). Discovering causal graph typically requires extra-statistical (beyond observational) information, for which the use of LLMs may be appropriate. The DAG can subsequently be used for variable selection when specifying a model, thereby avoiding the misspecification pitfalls mentioned. In the ensuing Q&A, Professor López de Prado answered questions on e.g. the time variation of DAGs and whether collider variables were important in forward looking models.
We thank Professor Marcos López de Prado and Nils Everling for their presentation.
Do you want to attend the next asset owner lunch of Inquire Europe? Reach out to Monique Sturmans (email: m.sturmans@inquire-europe.org)