ABAB Pedia

Louis Kirsch

Louis Kirsch is a researcher working on meta-learning and the automation of AI research who did his PhD with Jürgen Schmidhuber at IDSIA and worked as a research scientist at Google DeepMindGoogle DeepMindGoogle DeepMind is Google’s artificial intelligence research organization, formed in April 2023 by merging DeepMind, which Google acquired in 2014, with the Brain team from Google Research, and led by CEO Demis Hassabis. It developed AlphaGo, AlphaFold, and the Gemini family of models, and Hassabis and John Jumper received the 2024 Nobel Prize in Chemistry for AlphaFold’s protein structure prediction work.Open the full entry; he is now a co-founder of Inherent, a London-based AI research lab.

Contents9 sections
Key facts

Graduating from the Hasso Plattner Institute

According to his personal website, he graduated from the Hasso Plattner Institute in Germany as the top student of the class of 2017.⁠[1]

Completing a Master of Research at University College London

According to his personal website, he completed a Master of Research (MRes) in computational statistics and machine learning at University College London under David Barber, graduating as the top student of the class of 2018.⁠[1]

PhD Studies at IDSIA

From 2019 to 2023, he was a PhD student at IDSIA, the Swiss AI lab, supervised by Jürgen Schmidhuber; according to his personal website, he has defended his doctoral thesis, titled “Automating AI Research.”⁠[1]

In October 2019, he submitted a paper to arXiv with Sjoerd van Steenkiste and Jürgen Schmidhuber on improving generalization in meta reinforcement learning through learned objectives; according to his personal website, this was his first work on meta-learning reinforcement learning algorithms and was accepted at ICLR 2020.⁠[2][1]

His personal website also states that he was awarded a total of 550,000 GPU hours on the Swiss National Supercomputer.⁠[1]

Publishing a Paper on General-Purpose In-Context Learning

In December 2022, he submitted a paper to arXiv as first author with James Harrison, Jascha Sohl-Dickstein and Luke Metz, examining how meta-learning Transformers can yield general-purpose in-context learning.⁠[3]

Research Scientist at Google DeepMind

From 2024 to 2025, he was a research scientist at Google DeepMind; according to his personal website, he founded and led the research team there building LLM-based “AI Scientists.”⁠[1]

Co-Founding Inherent

According to his personal website, he co-founded Inherent in 2025 and serves as its “Chief Superintelligence Officer” (CΩO). Radical Ventures describes Inherent as a London-based research lab building general-purpose inventive AI to accelerate scientific discovery; its other co-founders are chief executive Tantum Collins, chief scientist Edward Hughes and chief technology officer Kaloyan Aleksiev.⁠[1][5]

Inherent’s website states that the company aims to recursively self-improve to discover new knowledge and that it is incorporated as a public benefit corporation.⁠[6]

Inherent Receives Seed Investment from Radical Ventures

Radical Ventures announced a seed investment in Inherent, and its portfolio page lists the investment year as 2026; the amount was not disclosed in the public materials.⁠[4][5]

Related Organizations, Websites, and Public Accounts

Personal website: Public link: https://louiskirsch.com/ (opens in a new window)⁠[6]

Sources

  1. Louis Kirsch (opens in a new window)
  2. Improving Generalization in Meta Reinforcement Learning using Learned Objectives (opens in a new window)
  3. General-Purpose In-Context Learning by Meta-Learning Transformers (opens in a new window)
  4. The Race for Recursive Self-Improvement (opens in a new window)
  5. Inherent (opens in a new window)
  6. Living within the Experiment (opens in a new window)