Seminar: Responsible Artificial Intelligence

The seminars aim at the independent exploration of a scientific topic based on some publications and a high-quality presentation of the topic both in written (report) and oral form (presentation and Q&A sessions). This seminar is dedicated to the discussion of selected topics in responsible artificial intelligence. Each trimester will focus on a different topic, e.g. multi-discrimination, post-hoc explainability, counterfactual explanations, explainability for certain types of data, etc.

• HT23 Focus: Responsible use of generative models

Generative models, such as Generative Adversarial Networks (GANs) and Variational Autoencoders (VAEs), have shown impressive capabilities in creating realistic data such as images, audio, and text, and are widely used in various fields: Image synthesis, style transformation, speech translation, text summarization, anomaly detection, game design, data enhancement, artistic design, healthcare and simulation of real-world scenarios for training purposes in fields such as aviation, healthcare and military. However, their power also raises ethical and societal issues that need to be addressed to ensure their responsible development and use. Key aspects of responsibility include ethical data use, privacy and security, transparency and explainability, bias and fairness, regulation and governance, accountability and attribution, and education and awareness.