L’intelligence générale artificielle : mirage ou faux raccord ?

Daniel Andler’s lecture offers a rigorous philosophical examination of the concept of Artificial General Intelligence (AGI). His argument is well-structured, moving from historical context to contemporary debates, and he carefully distinguishes between different definitions of AGI. The talk is notable for its intellectual honesty, as Andler openly admits his initial misunderstanding of ‘faux raccord’ and uses it as a springboard for deeper analysis. He presents a balanced view, acknowledging the impressive achievements of AI while questioning the assumption that these advances necessarily lead to AGI. The distinction between ‘mirage’ and ‘faux raccord’ is insightful: the former suggests that AI will never reach human-level intelligence, while the latter implies that AI may appear to approach it but only through a superficial resemblance. Andler’s argument that AI lacks an essential dimension of human intelligence—such as embodiment, social context, and intentionality—is compelling and well-supported by references to cognitive science and philosophy. He also addresses common counterarguments, such as the rapid progress of AI, but contends that these are based on a misunderstanding of the nature of intelligence. The lecture is not anti-AI; rather, it calls for a more nuanced understanding of what intelligence entails. The sources cited, including historical figures like Herbert Simon and Marvin Minsky, and contemporary debates, lend credibility to his argument. However, some claims could benefit from more detailed citations, and the lecture is more philosophical than empirical. The title accurately reflects the content, and the talk is a valuable contribution to the discourse on AGI. The audience’s reactions, as indicated by the comments, show a mix of appreciation for the depth of the analysis and some disagreement, but overall the lecture is well-received.

The title accurately reflects the content, which explores whether AGI is a mirage or a false continuity.

8/10

The speaker is a renowned philosopher and mathematician, and the talk is part of a prestigious academic colloquium. The arguments are well-structured and grounded in historical and contemporary references, though some claims are not fully sourced.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and acknowledgment of title error
• 03:00Historical predictions of AGI by Simon and Minsky
• 07:00Definition of AGI and two families of conceptions
• 10:00Thesis that AI tends toward AGI and its popularity
• 15:00Introduction of the mirage metaphor and series convergence example
• 20:00Arguments for the mirage: failures of LLMs and lack of common sense
• 25:00Transition to the false continuity argument
• 30:00Discussion of the essential dimension missing in AI
• 35:00Conclusion and implications for AGI
• 40:00Q&A session begins

• Collège de France - Colloque de rentrée 2025 — Official page of the colloquium where this talk was given.
• Collège de France — Institution's main website.
• Fondation du Collège de France — Support page for the institution.
• Threads - Collège de France — Social media presence.
• Bluesky - Collège de France — Social media presence.
• LinkedIn - Collège de France — Social media presence.

• Sébastien Bubeck et al., 'Sparks of Artificial General Intelligence: Early experiments with GPT-4' — Referenced in the talk as evidence of perceived AGI sparks.

• Ray Kurzweil's predictions on AGI — Kurzweil predicts AGI by 2029, contrasting with Andler's skepticism.