JavaScript is required to use this site. Please enable JavaScript in your browser settings.

Leading PI

Dr. Christoph Lehmann

At the 10th International Summer School on AI and Big Data, Dr. Christoph Lehmann will talk about Data, Assumptions, Models and Uncertainties: A Statistical View on Responsible AI.

Talk: Data, Assumptions, Models and Uncertainties: A Statistical View on Responsible AI

Responsible AI sounds promising, but its practical implementation remains largely nebulous. Key points for action are potentially around transparency and a robust understanding of the methodologies employed in data analytics, machine learning, and AI. Taking a statistical perspective, this talk aims to shed light on some not so obvious areas when employing methods from data analytics, machine learning, or AI.

Within this context, the first part refers to data-driven approaches. Data-driven approaches in data analytics and beyond are often favored for their assumption-free nature, allowing the data to speak for itself. But are these approaches truly devoid of assumptions? How do they differ from methods with more assumptions? Subsequently, the distinction between the estimation and prediction problems is explored. Machine learning and AI typically refers to prediction tasks, contributing to its notable success. Enhancing awareness of estimation versus prediction aids in method selection tailored to the specific problem at hand.

Lastly, uncertainties play a crucial role in experiment planning and analysis. While commonplace in natural sciences, uncertainty quantification progresses more slowly in machine learning and AI. This talk offers the perspective of considering the training of a machine learning model an uncertainty source in an experiment.

Bio

Christoph Lehmann received his PhD in Statistics from the Faculty of Business and Economics at TU Dresden, following his studies in Industrial Engineering. His initial research focused on financial markets and credit risk from 2007 to 2017. Since 2017, he has been a postdoctoral researcher at ScaDS.AI, where he works on quantifying uncertainties in machine learning and AI applications.
In addition to his statistical perspective, he is also interested in ethical questions related to the application of AI. Alongside his research, he has been advising companies on data analysis and has been a lecturer in Statistics at the University of Cooperative Education Dresden since 2016.

funded by:
Gefördert vom Bundesministerium für Bildung und Forschung.
Gefördert vom Freistaat Sachsen.