Tim Bakker

Tim Bakker

PhD researcher in Machine Learning

University of Amsterdam

Biography

Currently finishing my PhD at the Amsterdam Machine Learning lab with Herke van Hoof and Max Welling. My research has primarily focused on active learning and active sensing for high-tech applications, such as MRI and scientific simulations, typically coupled with reinforcement learning. I’ve completed research internships at Facebook AI Research (FAIR) and Qualcomm AI Research.

Other interests include AI safety, everything Bayesian, effective altruism, and musical theatre.

I am concerned about the risks of transformative AI, and have spoken on this at the Dutch National AI Debate, Pakhuis de Zwijger, the Dutch Ministry of Defence, and the University of Amsterdam.

Download my resume (generally not as up-to-date as this website).

Interests
  • Active learning / sensing
  • AI alignment
  • Bayesian probability theory
  • Musical theater
Education

  • MSc in Theoretical Physics, 2016

    University of Amsterdam

  • BSc in Physics and Astronomy, 2014

    University of Amsterdam

Talks and media appearances

Panel on the Existential Risks of AI with prof. Roman Yampolskiy
Panel on the existential risks of artificial intelligence, with Roman Yampolskiy, Simon Friederich, Jesse Six Dijkstra, Queeny Rajkowski, Marieke Koekkoek, Lisa Gotoh and Tim Bakker.
Thu, Jun 6, 2024 20:00 Pakhuis de Zwijger, Amsterdam
Video
Panel on the Existential Risks of AI with prof. Roman Yampolskiy
Nationaal AI Debat
Dutch national AI debate, where I will speak on the risks of AI together with Kees Verhoeven, Felienne Hermans, and Cynthia Liem.
Wed, Nov 1, 2023 19:00 — 20:30 Spijkenisse, Netherlands
Nationaal AI Debat
Academic Panel on AI Risks
Academic panel on the risks of artificial intelligence, with Iris Groen, Eric Nalisnick, Jakub Tomczak, and Tim Bakker. With a keynote by Ajeya Cotra.
Wed, Oct 18, 2023 18:00 — 20:00 Science Park 904, University of Amsterdam
Academic Panel on AI Risks
New York WHCR radio; AI with Tim Bakker
Radio appearance on Community & Technology on WHCR-FM 90.3, with hosts Stuart Reid and Dave Burstein. We discussed the risks and opportunities of artificial intelligence.
Wed, Oct 11, 2023 17:00 — 18:00 Online
Video
New York WHCR radio; AI with Tim Bakker
Learning objective-specific active learning strategies with Attentive Neural Processes
Presentation on our ECML 2023 paper “Learning objective-specific active learning strategies with Attentive Neural Processes”.
Thu, Sep 21, 2023 10:30 Polytechnic University of Turin
Learning objective-specific active learning strategies with Attentive Neural Processes
See all events

Publications

Tim Bakker, Thomas Hehn, Tribhuvanesh Orekondy, Arash Behboodi, Fabio Valerio Massoli (2023). Active Learning Policies for Solving Inverse Problems. NeurIPS ReALML Workshop 2023.

PDF Cite

Tim Bakker, Fabio Valerio Massoli, Thomas Hehn, Tribhuvanesh Orekondy, Arash Behboodi (2023). Switching policies for solving inverse problems. NeurIPS Deep Inverse Workshop 2023.

PDF Cite

Tim Bakker, Herke van Hoof, Max Welling (2023). Learning objective-specific active learning strategies with Attentive Neural Processes. ECML 2023.

PDF Cite Code

Teodora Pandeva, Tim Bakker, Christian A. Naesseth, Patrick Forre (2022). E-Valuating Classifier Two-Sample Tests. Preprint.

PDF Cite

Tim Bakker, Matthew Muckley, Adriana Romero-Soriano, Michal Drozdzal, Luis Pineda (2022). On learning adaptive acquisition policies for undersampled multi-coil MRI reconstruction. MIDL 2022.

PDF Cite Code

See all publications

Blog posts

Risky optimisation
Viewing AI risk through the lens of optimisation provides intuitions for why AI risk is a problem and for what potential solutions must achieve.
Tim Bakker
Last updated on Tue, Oct 24, 2023 19 min read
Risky optimisation
AI alignment
My views on the importance and difficulty of the AI alignment problem.
Tim Bakker
Last updated on Wed, Oct 18, 2023 32 min read
AI alignment
Priors and generalisation
A look at the role of priors in deep learning methods.
Tim Bakker
Last updated on Thu, Oct 19, 2023 9 min read
Priors and generalisation
From Bayesian inference to modern machine learning
A tour from prescriptive Bayesianism to modern statistical machine learning.
Tim Bakker
Last updated on Thu, Oct 19, 2023 14 min read
From Bayesian inference to modern machine learning
Everything but commentary
A look at the objective Bayesian prescription for reasoning under uncertainty.
Tim Bakker
Last updated on Wed, Oct 18, 2023 12 min read
Everything but commentary

Supervision

Master AI projects

  • 2021-2023: Active learning for imbalanced image classification, Philip Goto.
  • 2020-2021: Uncertainty and Diversity Based Methods in Batch-Mode Active Learning, András Csirik.
  • 2020-2021: Back to Basics: Deep Reinforcement Learning in Traffic Signal Control, Sierk Kanis. UrbComp'21 Best Paper Award runner-up.
  • 2019-2020: Contrastively-shaped reward signals for curiosity-based exploration tasks, Nil Stolt Anso.

Bachelor AI projects

  • 2019-2020: Forecasting shipped orders with long short-term memory recurrent neural networks, Roel van der Burght.

External projects

  • 2019-2020: Bayesian Convolutional Neural Networks for Learning Ground State Energies from a Simple Harmonic Potential Well, Chase van de Geijn.

Teaching

Reinforcement Learning 2019, 2020

I was teaching assistant for the Master AI Reinforcement Learning 2019 and 2020 course at the University of Amsterdam. This involved leading lab and theoretical homework sessions, as well as supervising group projects. Additionally, I taught a substitute lecture on Deep Q-Learning.

Leren (Introduction to Machine Learning) 2019, 2021

I was teaching assistant for the Bachelor AI Leren 2019 and 2021 course at the University of Amsterdam. This involved leading theoretical homework sessions, and assisting students in their lab work.

Experience

 
 
 
 
 
Research intern
Qualcomm AI
May 2023 – Sep 2023 Amsterdam
Research internship on reinforcement learning for efficient optimisation of inverse problems.
 
 
 
 
 
Research intern
Facebook AI Research
Jun 2021 – Oct 2021 Montreal (remote)
Research internship on learning adaptive subsampling in Magnetic Resonance Imaging (MRI).
 
 
 
 
 
PhD researcher
AMLab, University of Amsterdam
Mar 2019 – Present Amsterdam
Active learning and active sensing, reinforcement learning, Bayesian methods, and AI alignment.
 
 
 
 
 
Machine Learning Engineer
BrainCreators
Sep 2017 – Feb 2019 Amsterdam
Engineer and consultant on a variety to topics; from active learning and clustering to object detection and natural language processing.

Volunteering

 
 
 
 
 
Chair of the board
Amsterdams Studentenprojectkoor (ASPK)
Jul 2024 – Present Amsterdam
The ASPK is an Amsterdam-based amateur choir and theatre association that performs operettas in Dutch.
 
 
 
 
 
Advisory board member
AI Safety Amsterdam
Sep 2023 – Present Amsterdam
AI Safety Amsterdam is an organisation dedicated to promoting AI Safety, primarily in the academic environment at the University of Amsterdam. The goal is to create awareness and educate about the importance of AI Safety research.
 
 
 
 
 
Guide and organiser
Effective Altruism Meditation
Oct 2020 – May 2021 Online
EA Meditation worked to connect effective altruists with an interest in meditation, and give them a place to practice together.
 
 
 
 
 
Mentor
Inclusive AI
May 2019 – Present Amsterdam
IAI mentors are responsible for helping mentees navigate the academic and non-academic world during and after their Master AI programme.
 
 
 
 
 
Board member
Inclusive AI
May 2019 – Oct 2021 Amsterdam
The IAI is an inclusive space for members to receive career-oriented help from senior peers in the field and connect with people of various backgrounds.
 
 
 
 
 
Chair of the activities committee
Amsterdams Studentenprojectkoor (ASPK)
Oct 2018 – Jan 2024 Amsterdam
The ASPK activities committee is responsible for organising regular events for ASPK members with the aim of fostering group cohesion and fun.
 
 
 
 
 
Co-founder and organiser
LessWrong Netherlands meetup
Jul 2016 – Jul 2018 Amsterdam and Utrecht
The LessWrong community is dedicated to improving human reasoning and decision-making. Each day, we aim to be less wrong about the world than the day before.
 
 
 
 
 
Community organiser
Effective Altruism Netherlands
May 2016 – Jul 2017 Netherlands
Effective Altruism is a philosophy and global social movement that is dedicated to using reason and evidence to do the most good tackling pressing problems in the world, such as global poverty, animal welfare, and global catastrophic risk.
 
 
 
 
 
Founder and organiser
Effective Altruism Amsterdam
Mar 2016 – Present Amsterdam
Effective Altruism is a philosophy and global social movement that is dedicated to using reason and evidence to do the most good tackling pressing problems in the world, such as global poverty, animal welfare, and global catastrophic risk.

Contact