Raphaël Millière

Timeline

1. By 2020

   Philosophy degrees and an Oxford DPhil

   He holds a BA in philosophy from Sorbonne University and an MA in philosophy from the Institut Jean Nicod at École Normale Supérieure. In 2020, he completes his DPhil in philosophy at the University of Oxford.

2. 2020–2023

   Presidential Scholar at Columbia

   From 2020 to 2023, he is the Robert A. Burt Presidential Scholar in Society and Neuroscience at Columbia University, working in the Center for Science and Society. In September 2020, Nautilus publishes his essay "Welcome to the Next Level of Bullshit," subtitled "The language algorithm GPT-3 continues our descent into a post-truth world."

3. 2023

   Lecturer at Macquarie University

   He joins Macquarie University in Sydney in 2023 as a Lecturer (Assistant Professor) in Philosophy of Artificial Intelligence. In November 2023, he and Charles Rathkopf publish the Vox essay "Why it's important to remember that AI isn't human."

4. January 2024

   A philosophical primer on language models

   In January 2024, he and Cameron Buckner post "A Philosophical Introduction to Language Models, Part I." The paper serves as a primer on language models for philosophers and a survey of their significance for classic debates on compositionality, language acquisition, semantic competence, grounding and world models.

5. 2024

   AI2050 Early Career Fellowship

   He becomes a 2024 Early Career Fellow in the AI2050 program of Schmidt Sciences. His project aims to build a framework for understanding how AI systems process information by identifying and overcoming "interpretability illusions" that mislead us about their inner workings.

6. 2025

   Variable binding in Transformers

   In 2025, he coauthors "How Do Transformers Learn Variable Binding in Symbolic Programs?" with Yiwei Wu and Atticus Geiger, published at the Forty-second International Conference on Machine Learning. Causal interventions show that the trained Transformer uses the residual stream as an addressable memory space, with specialized attention heads routing information across token positions.

7. 2025

   The 2025 AI2050 Prize

   The American Philosophical Association awards him the 2025 AI2050 Prize for Early Career Researchers for his paper "Normative Conflicts and Shallow AI Alignment," published in Philosophical Studies in 2025. The paper shows how conflicts between norms such as helpfulness, honesty and harmlessness create vulnerabilities in language models that bad actors can exploit.

8. As of October 2026

   Associate Professor at Oxford

   He is an Associate Professor of Theoretical Philosophy at the University of Oxford and a Tutorial Fellow at Jesus College, with an affiliation at the Institute for Ethics in AI. His research examines the capacities and limitations of AI systems built on deep neural networks, and he holds an AI2050 Fellowship from Schmidt Sciences.