Regensburg 2025 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

SYAI: Symposium AI in (Bio-)Physics

SYAI 1: AI in (Bio-)Physics

SYAI 1.3: Hauptvortrag

Donnerstag, 20. März 2025, 10:30–11:00, H1

Emergent behavior of artificial intelligence — •Steffen Rulands — Arnold-Sommerfeld Center for Theoretical Physics, LMU München, Theresienstr. 37, 80333 München

Can artificially intelligent systems show emergent behavior that does not originate from the data they are trained on? This question is relevant for understanding the potential dangers and possibilities of powerful AI systems. In this talk, I will address this question from two different perspectives drawing on ideas from theoretical biophysics. First, I will show that neural networks exhibit an instability that gives rise to the emergent formation of structures independent of the training data. I will show how these structures aid the interpretation of data. In the second part of my talk, I will then draw on the biophysics of subcellular compartments and insect societies to demonstrate ways in which emergent behavior of artificially intelligent systems can facilitate and jeopardize deep learning.

Keywords: artificial intelligence; machine learning; biophysics