About Me
I am a Researcher in the Universal AI team at DeepMind, London (UK). Our team works on understanding the foundations of general intelligence, researching the mathematical and algorithmic foundations of sequential decision makers. I am interested in understanding the fundamental computational mechanisms underlying today's AI systems, and how to improve their safety and reliability. Most recently I got interested in the fundamentals of how to unambiguously tell AI systems what to achieve, and how to go about it (and how not to go about it; which touches upon questions of social norms and ethics). For a summary and pointers to more in-depth discussions on my blog see my Research pages.
Mini-CV
| Research Scientist, Universal AI Team, DeepMind. | |
| Research Scientist, Safety Analysis Team, DeepMind. | |
| Research Scientist, Deep Learning Perception group, Bosch Center for AI. | |
| PhD candidate, Sensorimotor Learning and Decision-Making group, Max Planck Institutes for Biological Cybernetics and Intelligent Systems. | |
| BSc & MSc, Telematics, Graz University of Technology. |
Links
- 🎓 Google Scholar for an up-to-date publication list
- 💼 [My CV] (coming soon)
- 🐙 GitHub
- 📨 tim܂genewein﹫gmail܂com (retype, don't copy&paste)
Academic service
I am an active action editor and reviewer for TMLR, area chair for UAI ('21, '22, '23, '24) and the AAAI Safe and Robust AI track ('23, '24), and a member of the ELLIS Society. I regularly review for ML conferences, including NeurIPS, ICLR, ICML, UAI, AAAI, AISTATS, and have received top reviewer certifications for NeurIPS ('18, '19, '20, '21), ICML ('19, '20), ICLR ('22) and TMLR (Expert reviewer, received in '23).
Past Affiliations and Topics
Safety Analysis Team at DeepMind: I was a founding member of DeepMind's Safety Analysis Team, led by Pedro Ortega, where I worked from 2018 to 2022. The team was part of the Technical AGI Safety group and focused on "understanding AI systems of today and the future" and how to build them safely. My research was focused on analysis and interpretability of AI agents. The Universal AI Team partly grew out of the Safety Analysis Team, and many of the research questions carried over.
Bosch Center for AI: Before joining DeepMind, I worked as a Research Scientist in the Deep Learning Perception group at the Bosch Center for AI (in Renningen, Germany) on Bayesian deep learning and neural network compression, and the combination of both.
PhD: 📥 Download Thesis
My PhD years were well spent in the Sensorimotor Learning and Decision-Making research group lead by Daniel A. Braun at the Max Planck Institute for Intelligent Systems and the Max Planck Institute for Biological Cybernetics in Tübingen, Germany. Our group was interested in connecting sensorimotor learning with theoretical models for inference and decision-making - with a particular focus on bounded-rationality. My research has been focused on understanding how decision-makers can leverage the statistical structure of their environment in order to efficiently cope with uncertainty. I investigated theoretical links between bounded rationality and information theory (rate distortion theory), and structure learning and structure selection mechanisms in humans through virtual reality experiments. Find more on my PhD research here.
Undergrad: I obtained my BSc and MSc in Telematics (now renamed into Information and Computer Engineering) at Graz University of Technology, Austria. My MSc Thesis was supervised by Gerhard Neumann and Wolfgang Maass.
Content Usage
Content on this site can be used for non-commercial purposes. Please be kind and attribute appropriately. If in doubt, just ask. I do not consent to any content on this site being used to train, fine-tune, test, or otherwise improve generative AI models. As of June 2024 no content on this site has been directly written by or edited with the assistance of an AI model.
Disclaimer
Content on this site may contain errors, and particularly the blog posts may express my opinions and speculative thought. Some of my views and opinions may change over time (as they should!). Do not quote me on decades-old blog posts. But do ask me whether my opinion has changed or not and why.
Contact
Find my current institutional email address by looking into one of my recent publications, or use tim܂genewein﹫gmail܂com (retype it, don't copy&paste).