Hello! I'm a PhD student in Computer Science at Umass Amherst advised by Ben Marlin. My research aims to reduce catastrophic risks from advanced AI systems.
I'm currently working on training weak models to decode the activations of strong models. I previously worked on learning diverse generalizations of underspecified data with applications to "measurement-tampering" (see below).
You can reach me at odanielskoch at umass dot edu.
Google Scholar • LessWrong • Twitter • GitHub •