Sessions

What makes something valuable? Is value a stable property of the world, or does it depend on the goals we are trying to achieve? This session will explore how value representations support flexible learning and decision-making in humans and AI.

This causerie session is motivated by recent news, and notably the controversy involving Anthropic vs the Pentagon. We will discuss power struggles in the race to control AI, as well as challenges in the global distribution of resources.

AI research evolves as a network of interacting ideas, communities, and technologies. Using phylo-epistemic networks, the talk explores the history of modelling language as a complex adaptive system. This case study highlights how alternative research trajectories can live, raising important questions about the diversity and future directions of AI research.

From the democratization of dual-use biotechnology to the emergence of deceptive agentic systems, the trajectory of Artificial Intelligence raises urgent questions about control and alignment. In this talk, Charbel-Raphaël Segerie presents a critical analysis of the current race toward superintelligence and the widening gap between technological capabilities and control mechanisms.

Large language models are rapidly entering classrooms, reshaping learning, assessment, and teaching practices. This session explores the ethical challenges raised by educational LLMs, drawing on the OpenLLM France 2030 project, and focuses on applied issues such as detecting LLM-assisted cheating and ensuring robustness against adversarial attacks in educational settings.

We review AI in 2025, make our bets for 2026, and see which predictions stand the test of time. Winner gets a prize!

How can we address the systemic challenges of the digital ecosystem? This festive session will be shaped by community discussion.

In this talk, Pierre Bessière presents the central ideas of his new book: why having a body fundamentally shapes intelligence, why disembodied AI systems depend on humans for meaning, and how natural agents remain necessarily ignorant of the world they act in. He contrasts two models of reasoning: logic and probability, arguing that only probabilistic reasoning allows intelligences to act meaningfully under uncertainty.

From creative studios to classrooms and boardrooms, generative AI is reshaping the way we work, learn, and imagine. This first Causerie of AI-Phi 3.0 invites an open discussion on the hopes and anxieties surrounding this technology: from the promise of universal creativity and efficiency to the spectre of runaway intelligence. Together with Edmund Ronald, we’ll reflect on how trust, safety, and understanding evolve as AI spreads through everyday life.

Dominique Lestel explores humanoid robots like Sophia as “Insurrectionary Machines” — artifacts that unsettle rather than serve, existing in friction with humans. From ancient idols to AI systems, these machines invite us to rethink life itself through an ethological lens.

Kathinka Evers explores why the question of artificial sentience (or “awareness”, or “consciousness”) is raised in science receiving considerable public funding despite the lack of empirical evidence? What this might mean and what the implications of it are.

Are qualitative differences in conscious experience, like vision versus audition, shaped by the intrinsic structure of neural networks? This talk proposes a mathematical framework using symmetry groups and equivariance to characterize these differences. This approach offers a new path for linking consciousness to neural architecture.

News and community updates session.

Can traditional notions of consent meaningfully apply to AI-generated content? What happens when our data creates representations of ourselves beyond our control?

How can we ensure AI systems' alignment with human values?

Short presentation(s) followed by collective discussions on reasoning in AI systems. We will celebrate this 20th session with a cake 🥳

In this informal session, we will cover some of the main innovations in LLMs since ai-phi 1.0.

A talk on Value-Aware Artificial Intelligence by Giulio Prevedello.

1st ai-phi year Closing Session

AI assertions by Merel Semeijn

Screening of Daniel Dennett's interview about consciousness.

Interactive discussion around Kahneman's interview with Lex Fridman

Generative AI, Sequence Models and Training Paradigms by Stefan Lattner (Sony CSL)

Introduction to Deep Learning by Stefan Lattner (Sony CSL)

Agency in Humans and in Machines by Lucas Battich and David Colliaux

Artificial Consciousness by Gaspard Fougea (ENS Saclay)

AI-Forum Session Discussion

Session on AEmotional Understanding by Constant Bonard and Gustave Cortal

X-mas Session and NeurIPS report by Stefan Lattner

Session on Explainability and Fairness by Thomas Souverain

Session on Effective Altruism by Jonathan Claybrough

The second half of the session on AGI. This focuses on different approaches to characterising AGI and finishes with an exploration of the "Sparks of AGI" paper.

Session on Artificial General Intelligence by Michael Anslow. This first session starts with a whirlwind introduction to Artificial Intelligence, focusing on search, knowledge representation and reasoning and machine learning.

Aïda Elamrani presented an Introduction to Philosophy Discussion

Session on Introduction to Philosophy by Aïda Elamrani