Large language models (LLMs) like GPT-4 or LlaMa 3 provide superior performance in complex human-like interactions. But they are costly, or too large for edge devices such as smartphones and harder to self-host, leading to security and privacy concerns. This paper introduces a novel interpretable knowledge distillation approach to enhance the performance of smaller, more economical LLMs that firms can self-host. We study this problem in the context of building a customer service agent aimed at achieving high customer satisfaction through goal-oriented dialogues. Unlike traditional knowledge distillation, where the “student” model learns directly from the “teacher” model’s responses via fine-tuning, our interpretable “strategy” teaching approach involves the teacher providing strategies to improve the student’s performance in various scenarios. This method alternates between a “scenario generation” step and a “strategies for improvement” step, creating a customized library of scenarios and optimized strategies for automated prompting. The method requires only black-box access to both student and teacher models; hence it can be used without manipulating model parameters. In our customer service application, the method improves performance, and the learned strategies are transferable to other LLMs and scenarios beyond the training set. The method’s interpretabilty helps safeguard against potential harms through human audit.

Typically, organizations lose around five percent of their revenue to fraud. In this presentation, we explore advanced AI techniques to address this issue. Drawing on our recent research, we begin by examining cost-sensitive fraud detection methods, such as CS-Logit which integrates the economic imbalances inherent in fraud detection into the optimization of AI models. We then move on to data engineering strategies that enhance the predictive capabilities of both the data and AI models through intelligent instance and feature engineering. We also delve into network data, showcasing our innovative research methods like Gotcha and CATCHM for effective data featurization. A significant focus is placed on Explainable AI (XAI), which demystifies high-performance AI models used in fraud detection, aiding in the development of effective fraud prevention strategies. We provide practical examples from various sectors including credit card fraud, anti-money laundering, insurance fraud, tax evasion, and payment transaction fraud. Furthermore, we discuss the overarching issue of model risk, which encompasses everything from data input to AI model deployment. Throughout the presentation, the speaker will thoroughly discuss his recent research, conducted in partnership with leading global financial institutions such as BNP Paribas Fortis, Allianz, ING, and Ageas.

The Sustainable Development Goals (SDGs) of the United Nations provide a blueprint of a better future by “leaving no one behind”, and, to achieve the SDGs by 2030, poor countries require immense volumes of development aid. In this work, we develop a causal machine learning framework for estimating heterogeneous treatment effects of aid disbursements that inform optimal aid allocation. We demonstrate the effectiveness of our method using data with official development aid earmarked to end HIV/AIDS in 105 countries, amounting to more than USD 5.2 billion. For this, we first show that our method successfully computes heterogeneous treatment-response curves using semi-synthetic data. Then, using real-world HIV data, we find that an optimal aid allocation suggested by our method could reduce the total number of new HIV infections compared to current allocation practice. Our findings indicate the effectiveness of causal machine learning to inform cost-efficient allocations of development aid that maximize progress towards the SDGs.

The digital twin paradigm integrates information obtained from sensor data, system physics models, as well as the operational and inspection/maintenance/repair history of a physical system or process of interest. As more and more data become available, the resulting updated model becomes increasingly accurate in predicting the future behavior of the system or process, and can potentially be used to support several objectives, such as safety, quality, mission planning, operational maneuvers, process control and risk management. This seminar will present recent advances in using Bayesian computational methods that advance the digital twin technology to support all these objectives, based on several types of computation: current state diagnosis, model updating, future state prognosis, and decision-making. All these computations are affected by uncertainty regarding system properties, operational parameters, usage, and environment, as well as uncertainties in data and prediction models. Thus, uncertainty quantification becomes an important need in system diagnosis and prognosis, considering both aleatory and epistemic uncertainty sources. The Bayesian methodology is able to address this need in a comprehensive manner and aggregate the uncertainty from multiple sources. A wide range of use cases such as additive manufacturing, aviation system safety, and power grid operations will be presented.