ELLIS unit Amsterdam announces 3 new AI PhD positions


The European Laboratory for Learning and Intelligent System (ELLIS) unit Amsterdam is delighted to announce three new ELLIS unit Amsterdam PhD positions. These projects will conduct research on artificial intelligence by machine learning with an international collaboration scheme by leveraging the ELLIS Society network, which the unit is part of. 

The ELLIS unit Amsterdam seeks to serve as a hub between the Amsterdam AI ecosystem spearheaded by the University of Amsterdam and other leading European AI ecosystems which expedites high-quality and collaborative research in machine learning or a related research area. Specifically, the ELLIS unit Amsterdam’s PhD programme aims to foster research collaboration through joint supervision: one supervisor from the University of Amsterdam, and the other one from a different university affiliated with the ELLIS Society. 

The call for PhD proposals was opened internally for the second time in February 2023 which targeted Faculty Members of the ELLIS unit Amsterdam. The received proposals were ranked by taking into consideration, among others, research excellence, the added value of the collaboration, and contribution to diversity and inclusion.

More information on the ELLIS unit Amsterdam’s programmes can be found:


The awarded projects are:

Enhancing video-AI robustness via human representational alignment

State-of-the-art deep neural networks developed for video-AI have become very successful on tasks such as action recognition. However, like other forms of deep learning, contemporary video-AI suffers from problems with robustness to noise and generalization across domain shifts, which are ubiquitous in real-world settings. Current approaches to increase robustness mainly focus on scaling up the training data and model size, requiring increasing computational resources, which runs counter to the societal need to develop efficient and sustainable AI. Here, we propose to explore a radically different strategy towards enhancing video-AI robustness and generalization: aligning video-AI models with humans, using human ratings of perceived video similarity and human brain measurements.

The project is a collaboration between Dr. Iris Groen (ELLIS Member, University of Amsterdam) and Prof. Dr. Hildegard Kühne (ELLIS Member, Institut für Informatik, Goethe University Frankfurt and University of Bonn), and Prof. Gemma Roig (Goethe University Frankfurt).

Dynamics of Indirect Reciprocity in Human-AI Hybrid Populations

Understanding cooperation is a fundamental research quest across disciplines. On the one hand, explaining the evolution of cooperation is a key scientific question in the realm of evolutionary biology. On the other hand, from climate change mitigation to pandemics’ control, meeting today’s major societal challenges requires understanding cooperation dynamics. With the advent of Artificial Intelligence, studies on cooperation found a new application domain: How to design systems where artificial agents positively impact human cooperation? This PhD project will combine theoretical analysis (based on Evolutionary Game Theory, and Multi-agent Reinforcement Learning) and human-agent interaction experiments to understand how learning agents impact human dynamics of cooperation and indirect reciprocity.

The project is a collaboration between Dr. Fernando P. Santos (ELLIS Member, Institute of Informatics, University of Amsterdam) and Prof. Ana Paiva (ELLIS Fellow, Instituto Superior Técnico, University of Lisbon).

4D Perception of Interacting Humans from Videos

A long-standing goal of Computer Vision is to understand human actions from videos. Given a video, people effortlessly figure out what objects exist in it, the spatial layout of objects, and the pose of humans. Moreover, they deeply understand the depicted action. What is the subject doing? Why are they doing this? What is their goal? How do they achieve this? To empower computers with the ability to infer and exploit such abstract concepts from pixels, we need to devise appropriate datasets and algorithms for 3D perception. Since humans live in a 3D world, their physical actions involve interacting with objects. Think of how many times per day one walks to the kitchen, grabs a cup of water, and drinks from it. This involves contacting the floor with the feet for walking, contacting the cup with the hand for grasping and lifting it, and drinking from the cup while the mouth comes in contact with it, and maintains contact across time. Thus, to understand human actions, we need to reason in 3D (or 4D with the extra dimension of time) about humans and objects jointly.

The project is a collaboration between Dr. Dimitrios Tzionas (ELLIS Member, Institute of Informatics, University of Amsterdam) and Prof. dr. Otmar Hilliges (ELLIS Fellow, Department of Computer Science, ETH Zurich).

More information on the ELLIS unit Amsterdam’s programmes can be found: