Erik Jenner

CS PhD student

UC Berkeley

Hi there!

I’m a Computer Science PhD student at UC Berkeley working on how to align powerful future AI systems with human values, to reduce the existential risk posed by AI. I am advised by Stuart Russell and part of the Center for Human-Compatible AI. I am grateful to be supported by fellowships from the Future of Life Institute and Open Philanthropy.

Interests

AI existential safety
Reward learning

Education

PhD in Computer Science, since 2022

UC Berkeley
MSc in Artificial Intelligence, 2022

University of Amsterdam
BSc in Physics, 2020

Heidelberg University

Recent Posts

Distributions Part II: What can we do with distributions?

As promised in part I, we can do a lot of the same things with Schwartz distributions as with classical functions. To see how, we’ll cover derivatives, convolutions, and Fourier transforms of distributions.

Mar 6, 2022 7 min read

Visualizing random vs grid search

Random search usually works better than grid search for hyperparameter optimization. This brief post suggests a way to visualize the reason for this geometrically.

Dec 22, 2021 2 min read

Extensions of Karger's algorithm

If you prefer videos, check out our ICCV presentation, which covers similar content as this blog post. For more details, see our paper. Karger’s contraction algorithm is a fast and very famous method for finding global minimum graph cuts.

Sep 10, 2021 8 min read

See all posts

Featured Publications

Erik Jenner, Enrique Fita Sanmartín, Fred A. Hamprecht

October 2021 ICCV (Oral)

Extensions of Karger's Algorithm: Why They Fail in Theory and How They Are Useful in Practice

We prove impossibility results showing that Karger’s contraction algorithm cannot be extended to $s$-$t$-mincuts or normalized cuts. However, we show how extensions of Karger’s algorithm can still be useful for seeded segmentation.

Erik Jenner, Adam Gleave

September 2021 NeurIPS Cooperative AI workshop

Preprocessing Reward Functions for Interpretability

We present a method for simplifying a learned reward model before visualizing it and show that this can make the reward more interpretable.

Erik Jenner, Maurice Weiler

June 2021 ICLR 2022

Steerable Partial Differential Operators for Equivariant Neural Networks

We present a framework for equivariant partial differential operators, generalizing existing approaches and narrowing the gap between PDOs and convolutions.

Research Experience

Master’s thesis

University of Amsterdam / UC Berkeley

Nov 2021 – Jul 2021 Amsterdam

Worked on the theory of potential shaping and on distance measures between reward functions with Adam Gleave (UC Berkeley) and Prof. Herke van Hoof (UvA)

Research Intern

Center for Human-Compatible AI

Jul 2021 – Sep 2021 UC Berkeley (virtual)

Worked on interpretability of reward models together with Adam Gleave
Developed a method to simplify a given reward model while keeping it equivalent to the original
Published results at the NeurIPS Cooperative AI workshop
Implemented preference comparisons for the imitation library