2025 SOA ImpACT Conference:Session 3I: Building Trust in AI Generated Data for Actuarial Models and Other AI Applications-11.04.25 Recording

Actuaries and statisticians, with our deep understanding of data and modeling, are uniquely equipped to critically assess artificial intelligence, particularly in leveraging generative AI in workflows. Abundant skepticism toward AI persists in actuarial circles due to ethical and reliability concerns, and this caution is essential for advancing AI ethics. This presentation explores artificial intelligence beyond the pale of Large Language Models (LLMs), highlighting the use of synthetic data in actuarial modeling, emphasizing trust dynamics in model outputs. We will highlight recent advances in the 'second generative AI wave,' drawing parallels with innovations in robotics and autonomous vehicles, to demonstrate how these developments can enhance model accuracy and reliability while addressing ethical challenges. This session is jointly presented by Matt Nash, an AI expert and Andrew Sykes, a Kellogg Professor and former actuary. Key Learning Objectives: Evaluate Trust in AI Outputs: Explore methods to assess and ensure the trustworthiness of models trained on or operating within a strictly regulated environment, focusing on transparency, bias mitigation, and explainability. Understand Synthetic Data Applications: Learn how synthetic AI-generated data can be effectively integrated into actuarial models to improve predictive accuracy and address data scarcity or privacy concerns. Explore Advances in Generative AI: Gain insights into the imminent 'second generative AI wave,' including parallels with robotics and autonomous vehicles, and its potential to transform actuarial practices with ethical and reliable solutions.