Associate Professor in Geometric Deep Learning
Based at AMLab
(University of Amsterdam)
Leading the Ideal Machine Intelligence research area within AMLab. I work on Geometric Deep Learning, embedding the symmetries and structures of nature into the learning process. I am also a Research Fellow at New Theory and Director for the ELLIS Program on Geometric Deep Learning.
I am an Associate Professor in Geometric Deep Learning at the University of Amsterdam (AMLab). My research focuses on grounding artificial intelligence in the principles of physics and geometry, so that the representations it learns respect the structure of the world they describe.
Beyond my academic role, I serve as a Research Fellow at New Theory and as Director for the ELLIS Program on Geometric Deep Learning.
Before joining the UvA, I worked as a post-doctoral researcher in applied differential geometry at the Technical University Eindhoven (TU/e). I completed my PhD in Biomedical Engineering (cum laude) at TU/e, where I developed medical image analysis algorithms based on sub-Riemannian geometry in the Lie group SE(2)—work inspired by the mathematical principles underlying human visual perception.
I am honored to have received several recognitions, including the MICCAI Young Scientist Award 2018 and two personal research grants from the Dutch Research Council (NWO): a VENI grant (2019) on Context-Aware AI and a VIDI grant (2023) for the project SIGN (Scalable Inference of Geometry-Grounded Neural Representations).
Most of my work is on geometric deep learning, but I also think about what the systems we build are. With AI that talks and behaves more and more like a person, the question of machine consciousness comes up whether we like it or not. How people answer it depends a lot on their metaphysics, usually one they grew up with rather than chose, and for most of us in science that is physicalism. I came to an idealist view, and I think it's at least as reasonable a place to start. I wrote down how I got there in A Minimal Metaphysics.
From there, Anna Ciaunica and I argue in a position paper at ICML 2026 that AI is best treated as technology, and that under the metaphysics we argue to be most plausible, a system which seems sentient is but a functional mimic. It reproduces what a mind does, but there is no subject undergoing it. If we extend moral standing to such systems, I worry the grounds on which we extend it to each other get looser. I don't expect everyone to share my starting point. What I'd like people to ask themselves is the reverse question. If you think a machine could be conscious, what in your picture of the world makes that plausible?
PhD in Biomedical Engineering (cum laude) from TU/e. Formerly a post-doc in applied differential geometry. My roots are in the mathematics of sub-Riemannian geometry and visual perception.
NWO VIDI (2023): SIGN — Scalable Inference of Geometry-Grounded Neural Representations.
NWO VENI (2019): Context-Aware AI.
MICCAI Young Scientist Award (2018).
"Nearly all data is rooted in our physical world." Symmetry is foundational to physics, and I think this extends to the rest of nature, mind included. So it is not only our models of physics that should respect it, but also our models of intelligence.
Artificial Intimacy — "What do AI relationships tell us about the state of humanity?", an in-conversation piece with philosopher and cognitive scientist Anna Ciaunica (self-consciousness, embodiment, and human–AI interaction), recorded for Sophia Aryan's consciousness-focused interview series Consciousness Experience: watch on YouTube.
Giant's Shoulder podcast (March 23, 2026), a conversation on consciousness, representation, and whether machines can have minds: watch on YouTube.
Find me on X (@erikjbekkers) and Bluesky, or reach me by email.