You can leave anonymous feedback here and download a pdf version of the CV here.

Marius Hobbhahn

Twitter
Google Scholar

Work Experience

Apollo Research

May 2023 - now: Co-founder and CEO of Apollo Research

Epoch

June 2022 - April 2023: part-time research fellow for Epoch

Education

International Max-Planck Research School Tübingen, Germany

PhD with Philipp Hennig, 2020-2024
Topic: Making Bayesian ML fast and scalable

University of Tübingen, Germany

M.Sc., Machine Learning, 2018-2020

University of Tübingen, Germany

B.Sc., Computer Science, 2016-2019

University of Tübingen, Germany

B.Sc., Cognitive Science, 2015-2018

Mentorship and Academic Appointments

  • SPAR23 mentor (2023): Mentored 7 AI safety students in mechanistic interpretability
  • Reviewer (2022-now): For top Machine Learning Conferences, including NeurIPS and ICML
  • Lead organizer (2021): Of the ELLIS doctoral symposium
  • Supervisor (2020-2022): For 3 Bachelor and 1 Master thesis
  • Teaching assistant (2016-2022): For Probabilistic ML, Data Literacy and Math summer course

Selected Awards and Engagements

  • Pacific Economic Cooperations Councel (2024): AI workshop Speaker
  • Beijing Academy of AI (BAAI) (2024): Speaker at evaluations panel
  • AI safety summit (2023): Participant at UK AI Safety Summit at Bletchley Park
  • Emergent Ventures (2022): Award Grantee. 15 awardees per cohort; Grant for AI safety research
  • Best speaker (2020): German Debate Championships
  • Scholarship (2016): German Academic Scholarship Foundation

Skills

  • Languages English (C2), German (native), French (B2)
  • Coding Python
  • Machine Learning Pytorch > JAX

Selected Publications

Large Language Models can Strategically Deceive their Users when Put Under Pressure

Jérémy Scheurer, Mikita Balesni, Marius Hobbhahn (2024).
[arxiv]

Black-box access is insufficient for rigorous ai audits

Stephen Casper, Carson Ezell, Charlotte Siegmann, Noam Kolt, Taylor Lynn Curtis, Benjamin Bucknall, Andreas Haupt, Kevin Wei, Jérémy Scheurer, Marius Hobbhahn, Lee Sharkey, Satyapriya Krishna, Marvin Von Hagen, Silas Alberti, Alan Chan, Qinyi Sun, Michael Gerovitch, David Bau, Max Tegmark, David Krueger, Dylan Hadfield-Menell (2024). FAccT.
[paper]

Compute Trends Across Three Eras of Machine Learning

Jaime Sevilla, Lennart Heim, Anson Ho, Tamay Besiroglu, Marius Hobbhahn, Pablo Villalobos (2022). IJCNN.
[arxiv] [Our World in Data]

Will we run out of data? Limits of LLM scaling based on human-generated data

Pablo Villalobos, Anson Ho, Jaime Sevilla, Tamay Besiroglu, Lennart Heim, Marius Hobbhahn (2024).
[arxiv]

Laplace Matching for fast Approximate Inference in Generalized Linear Models

Hobbhahn, Marius and Hennig, Philipp. (2021).
[arxiv] [code] [blog]

Fast Predictive Uncertainty for Classification with Bayesian Deep Networks

Hobbhahn, Marius; Kristiadi, Agustinus and Hennig, Philipp. UAI 2022.
[arxiv] [code] [Blog post]