3rd Symposium on Machine Learning & Algorithmic Information Theory (AIT)

[boumediene hamzi]

Dear all,

Following the success of the first two editions - hosted at the Alan Turing Institute (2022), https://sites.google.com/site/boumedienehamzi/symposium-on-algorithmic-information-theory-and-machine-learning, and Imperial College London (2025), https://sites.google.com/site/boumedienehamzi/second-symposium-on-algorithmic-information-theory-and-machine-learning - we are pleased to announce the upcoming 3rd Symposium on Machine Learning and Algorithmic Information Theory, with a new emphasis on AI safety applications.

The symposium continues to explore the rich and growing interface between Algorithmic Information Theory (AIT) and Machine Learning (ML), bringing together researchers working across both theory and practice.

🎯 Objective

We aim to convene researchers at the intersection of:
*Algorithmic Information Theory
*Machine Learning
*Learning Theory
*Dynamical Systems
*Control Theory
*Agent Foundations
*AI Safety

In particular, we are interested in applying concepts such as Kolmogorov complexity and algorithmic probability to contemporary questions surrounding:
*AI generalization
*Agent behaviour
*Alignment and safety

🔎 Topics of Interest (including but not limited to)
1️⃣ Using AIT to explain a priori why certain ML methods succeed (or fail)
2️⃣ Theoretical foundations and practical applications of Solomonoff induction
3️⃣ Connections between information theory, learning, data compression, and agency
4️⃣ ML algorithms for improved prediction and compression approximating Kolmogorov complexity
5️⃣ Probabilistic modeling via algorithmic probability frameworks
6️⃣ Embedded agency and imprecise probability
7️⃣ Control-theoretic and dynamical systems perspectives on stability, robustness, and safety of learning algorithms

📍 Location: Mathematical Institute, University of Oxford
🗓 Timing: July
⏳ Duration: 2–3 days (depending on the number of speakers)

Before finalizing logistics - including securing the venue - we would like to gauge interest from potential speakers and participants.
If you are interested in contributing or attending, please complete this short form:
👉 https://docs.google.com/forms/d/e/1FAIpQLSfQd4iB0Vu1Iw-3NUIXHB0W8LQhcytbxjnJBTo-8jSOVgysoA/viewform?uusppreview

Your input will help shape the scope and structure of the event.

We look forward to fostering another stimulating and forward-looking discussion at the intersection of theory, learning, and AI safety.

Thanks and best wishes,

Boumediene