Louis Kirsch

Recording Slides Invited talk: Escape Velocity ICLR Recursive Workshop 2026 We are at an inflection point where recursive self improvement (RSI) transitions from proof of concept to take-off.

Paper Towards General-Purpose In-Context Learning Agents NeurIPS Workshops 2023 - GenPlan (contributed talk), FMDM, R0-FoMo, DistShift Reinforcement Learning (RL) algorithms are usually hand-crafted, driven by the research and engineering of humans.

Paper on arXiv General-Purpose In-Context Learning by Meta-Learning Transformers (MemARI & Meta-Learn NeurIPS Workshops 2022) Modern machine learning requires system designers to specify aspects of the learning pipeline, such as losses, architectures, and…

Paper on arXiv Eliminating Meta Optimization Through Self-Referential Meta Learning (DARL & AutoML Workshops 2022) Meta Learning automates the search for learning algorithms.

ArXiv page on SymLA Introducing Symmetries to Black Box Meta Reinforcement Learning (AAAI 2022) Meta reinforcement learning (RL) attempts to discover new RL algorithms automatically from environment interaction.

This year’s NeurIPS 2021 I present one full paper and a workshop paper on meta learning gradient-free & general-purpose learning algorithms.

Invited talk abstract (NeurIPS 2020 Meta Learning Workshop) Humans develop learning algorithms that are incredibly general and can be applied across a wide range of tasks.

Abstract (tldr) Biological evolution has distilled the experiences of many learners into the general learning algorithms of humans.

In September, I published a technical report on what I consider the most important challenges in Artificial Intelligence.

Deep Neural networks have rid us of feature engineering.