AI through the looking glass:
Interdisciplinary insights into challenges and risks
Interdisciplinary insights into challenges and risks
Registration
You will receive the Zoom link to follow the seminar remotely, together with the announcements of the talks and further updates, only after registering through the following form.
Aims and scope
The idea of setting up this interdisciplinary seminar emerged from the observation that the current development of Artificial Intelligence (AI) goes far beyond technological innovation. AI is rapidly transforming scientific working methods, research processes, higher education, knowledge transfer and many other aspects of our society at a pace that requires continuous dialogue across disciplines. Many of the questions arising from these developments can be adequately addressed only through collaboration between mathematics, computer science, the natural sciences, the social sciences, and the humanities.
With this seminar series, we aim to create an international forum based in the Berlin-Brandenburg area that regularly brings together researchers from different disciplines and enables a scientifically grounded, interdisciplinary exchange on the latest developments and their long-term consequences.
Contact
ai.and.the.looking.glass (at) gmail.com
Organisers
Hanno Gottschalk (TU Berlin)
Sylvie Paycha (University of Potsdam)
Elke Rosenberger (University of Potsdam)
Christoph Stephan (University of Potsdam)
Fabrizio Zanello (University of Potsdam)
Forthcoming sessions
Note: The titles, abstracts and slides of the previous talks are available on the "Previous talk" page of this website.
Wednesday, November 25th, TU Berlin, IMoS
4: 15 pm - Wojciech Samek (Fraunhofer Institute, Heinrich-Hertz-Institut and Technische Universität Berlin)
Title: The Science of Understanding AI: From Foundation Models to Autonomous Agents
Abstract: Artificial intelligence is rapidly evolving from predictive models to powerful foundation models and increasingly autonomous agents capable of reasoning, planning, and acting. While these advances create exciting opportunities, they also raise fundamental questions about how AI systems work, make decisions, and can be trusted. In this talk, I will highlight recent advances in explainable AI and mechanistic interpretability that help uncover the internal representations and mechanisms underlying AI behavior. I will discuss how these insights can support the monitoring, evaluation, and control of increasingly complex AI systems, and outline key challenges on the path toward reliable and trustworthy autonomous AI.
Friday, December 4th, University of Potsdam, Mathematics Institute
10am - Zoran Nikoloski (University of Potsdam)
Title: From sequences to systems: Bioinformatics reimagined by Artificial Intelligence
Abstract: Artificial intelligence has already made enormous advances in resolving key problems in fundamental and applied biology research, from protein structure to variant effect prediction. Yet, there remain significant challenges in understanding how certain phenotypes emerge in a particular biological system—knowledge that is essential for driving the system towards a desired state. In this talk, I argue that AI breakthroughs in biology strongly depend on the available data and the context-dependence of the addressed problem. I will also provide examples of which knowledge gaps AI have managed to fill and the path forward towards addressing the grand challenge of mechanistic understanding of multi-scale biological systems.
Wednesday, December 9th, TU Berlin, IMoS
4pm - David Schlangen (University of Potsdam)
Title: TBA
Abstract: TBA
Venues
Technical University Berlin
Wednesdays 4 - 6 pm (every other week)
IMOs, Fasanenstr. 89, 10623 Berlin, IMoS, Room 0004
(entrance at Müller-Breslau-Straße 10)
and
University of Potsdam , Golm Campus
Fridays 10 - 12 am (every other week)
Mathematics Institute, Building 9
Karl-Liebknecht-Str. 24-25
D-14476 Potsdam OT Golm