Tegmark Group
About
Updated 05/18/26The Tegmark AI Safety Group is an academic research lab at the Massachusetts Institute of Technology led by Professor Max Tegmark, a Swedish-American physicist and machine learning researcher. Tegmark joined MIT's Department of Physics in 2004 after earning tenure at the University of Pennsylvania. During his first 25 years as a researcher, he focused on cosmology and quantum information, but in the 2010s his research shifted toward machine learning and AI safety. The group's primary research focus is on mechanistic interpretability: given a trained neural network that exhibits intelligent behavior, how can we figure out how it works, preferably automatically? The group uses tools from physics and information theory to transform black-box neural networks into more understandable systems. Recent applications have included auto-discovery of symbolic formulas and invariants, as well as hidden symmetries and modularities in neural networks. A major contribution from the group was the development of Kolmogorov-Arnold Networks (KANs) in 2024, led by Tegmark and his PhD student Ziming Liu. KANs replace fixed activation functions on nodes with learnable activation functions on edges, providing dramatically improved interpretability for certain deep learning applications. The KAN paper became one of the most discussed ML papers of 2024. The group also conducts research on provably safe and guaranteed safe AI. Tegmark co-authored influential papers on this topic, including "Provably safe systems: the only path to controllable AGI" (2023) and "Towards Guaranteed Safe AI" (2024, co-authored with David Dalrymple, Yoshua Bengio, Stuart Russell, and others). This framework proposes AI systems equipped with high-assurance quantitative safety guarantees through world models, safety specifications, and formal verifiers. Tegmark is also the co-founder and president of the Future of Life Institute (FLI), a separate nonprofit focused on reducing global catastrophic risks from advanced technologies. He additionally founded the Beneficial AI Foundation (2022), a 501(c)(3) nonprofit supporting technical AI safety research on red-teaming, interpretability, and guaranteed safe AI. These are distinct organizations from the MIT research group. Tegmark is the author of over 300 publications and two bestselling books: "Our Mathematical Universe" (2014) and "Life 3.0: Being Human in the Age of Artificial Intelligence" (2017). He is a Fellow of the American Physical Society, received the Royal Swedish Academy of Engineering Science's Gold Medal in 2019, and was named one of Time Magazine's 100 Most Influential People in AI in 2023. The group is affiliated with the NSF AI Institute for Artificial Intelligence and Fundamental Interactions (IAIFI), the Center for Brains, Minds and Machines (CBMM), and the MIT Kavli Institute for Astrophysics and Space Research.
Theory of Change
The Tegmark AI Safety Group aims to reduce catastrophic risks from advanced AI by advancing mechanistic interpretability and frameworks for guaranteed safe AI. By studying how powerful models work internally and developing formal approaches that equip AI systems with provable safety guarantees, the group seeks to make future AI deployments more transparent, controllable, and aligned with human values, while training researchers who can apply these methods across high-stakes domains.
Community Signal
Updated 05/18/26Endorsements support Massachusetts Institute of Technology.
No endorsements yet.
Discussion
No comments yet. Be the first to share your thoughts.
Details
- Start Date
- -
- End Date
- -
- Expected Duration
- -
- Funding Raised to Date
- -