PIERRE BESSIÈRE - "Intelligences : artificielles et insensées ou naturelles et ignorantes"

PIERRE BESSIÈRE - "Intelligences : artificielles et insensées ou naturelles et ignorantes"

🎙 Pierre Bessière 👥 3K 📅 May 4, 2026 ⏱ 85 min 👁 107 📄 science communication 🧭 2026-08-02
Available in: English (current) Français

Keywords

embodied cognitionartificial intelligencenatural intelligenceBayesian inferenceChlamydomonas

Summary

In this conference, Pierre Bessière explores the diversity of intelligences, from artificial systems like Deep Blue and ChatGPT to natural ones like algae and species. He argues that having a body fundamentally changes intelligence, making problems more complex but also providing solutions. Through a series of duels, he illustrates that AI excels in closed, well-defined problems (like chess) but fails in open, physical-world tasks (like moving chess pieces). He contrasts the dexterity of a pianist with a gamer, and the adaptability of a rugby player with a bipedal robot. He also discusses the intelligence of unicellular organisms like Chlamydomonas, which can navigate towards light, and the collective intelligence of species. He concludes that natural intelligences are embodied and robust, while AI is disembodied and fragile, and suggests that Bayesian inference offers a promising framework for understanding and building adaptive systems.

140 words

Critical Evaluation

The talk provides a compelling and accessible overview of the differences between artificial and natural intelligence, emphasizing the importance of embodiment. Bessière’s arguments are well-structured, using vivid examples and analogies to illustrate complex concepts. The comparison between Deep Blue and a human assistant highlights the distinction between solving a closed problem and interacting with the physical world. Similarly, the contrast between Atlas and a rugby player effectively demonstrates the limitations of current robotics in adapting to unpredictable environments. The discussion of Chlamydomonas and species-level intelligence broadens the perspective, showing that intelligence exists at various scales. However, the talk lacks depth in some areas; for instance, the mechanisms underlying Bayesian inference in robots are only briefly mentioned. Additionally, while the speaker is credible, the absence of specific citations or references to studies weakens the scientific rigor. The title is well-matched, and the content is engaging, but the technical level is moderate, making it suitable for a general audience. Overall, the talk offers valuable insights into the nature of intelligence, but it could benefit from more detailed evidence and references.

178 words

Title / Content Match

The title accurately reflects the content, which contrasts artificial and natural intelligences, emphasizing the role of embodiment.

Quality & Reliability

7/10

The speaker is a recognized researcher in cognitive science, and the talk is grounded in established concepts (e.g., embodied cognition, Bayesian inference). However, the presentation is largely anecdotal and lacks detailed citations or references to specific studies, which limits its verifiability.

Key Moments

Cited Sources

  • Cognivence website — Organization of the Forum des Sciences Cognitives, where this talk was given.

Concurring Sources

Dissenting Sources

  • No discordant sources found — The talk does not contradict established scientific consensus; it aligns with mainstream views on embodied cognition and AI limitations.

Contribution & Novelties

The talk offers a fresh perspective on the comparison between artificial and natural intelligence, emphasizing the crucial role of embodiment. It synthesizes examples from robotics, biology, and AI to argue that natural intelligences are inherently adaptive and robust, while current AI systems are brittle and limited to closed domains. The speaker also introduces the concept of species-level intelligence, which is often overlooked.

Pour aller plus loin :

  • Embodied cognition — A foundational concept in cognitive science that the talk heavily relies on.
  • Bayesian inference — The probabilistic framework mentioned as a basis for adaptive behavior in robots.
  • Chlamydomonas — The unicellular alga discussed as an example of simple intelligence.
  • Deep Blue (chess computer) — The IBM computer that defeated Kasparov, illustrating AI’s strength in closed problems.

126 words

Radar Profile

The radar profile shows high scores in quantity and quality of information, moderate technical level, and good reliability. This indicates a well-balanced talk that is informative and credible, though not highly technical.

Reliability 7/10