Changeux, Jean-Pierre & Hakim, Vincent & da Silveira, Rava Azeredo

Changeux, Jean-Pierre: neurobiologist, Institut Pasteur and Collège de France, known for allostery and the global neuronal workspace model

Hakim, Vincent: theoretical physicist at the École normale supérieure, working on neural networks with Nicolas Brunel

da Silveira, Rava Azeredo: theoretical physicist at the École normale supérieure, working on the retina

Despite the title, Changeux's own lecture is lost: the recording opens on a short introduction, followed by a stretch of music and garbled captions. What remains are two English-language talks by physicists, identified from their collaborators and the speakers' references to "Vincent" and "Rava", and a discussion in which Changeux answers questions. The chair is not named.

Comment fonctionne le cerveau — Jean-Pierre Changeux

Andrea Cirla · Feb 23, 2025 · partial coverage

• Jean-Pierre Changeux
• Vincent Hakim
• theoretical neuroscience
• cerebellum
• silent synapses
• neural oscillations
• retina
• receptive fields
• global workspace
• physics and biology

Coverage note: partial - Changeux's talk is unreadable in the automatic captions; the two following talks and the discussion are covered in full.

Core ideas

Vincent Hakim asks what theoretical physics, beyond its experimental tools (NMR, microscopy, single-molecule manipulation), can bring to biology, and answers with two examples. The first concerns the synapses between parallel fibres and Purkinje cells in the cerebellum. Philippe Isope and Boris Barbour had measured several hundred of these synapses and found that more than 80% were silent. Taking David Marr's 1969 theory of the cerebellum as a learning machine, Hakim's team modelled the Purkinje cell as a simple threshold unit that classifies inputs, using Elizabeth Gardner's methods. If it learns as many associations as it can, more than half of its excitatory synapses end up at zero weight, because weights cannot go negative. This offers a new explanation for silent synapses, perhaps valid beyond the cerebellum. The second example concerns fast brain rhythms, as seen during sleep or in hippocampal slices, in which individual neurons fire irregularly without being oscillators. This differs from Huygens's classic picture of coupled clocks. A random network of simple inhibitory neurons with synaptic delays reproduces such collective rhythms. Their frequency is set mainly by synaptic transmission delay, which provides a basis for more detailed models.

Rava da Silveira presents the retina, a piece of brain tissue that sends the brain not a photocopy of the world but an actively processed message. Light intensity varies hugely, from about one photon per cell per minute under starlight to a hundred million at noon. The retina copes by sending differences between neighbouring points through "centre-surround" receptive fields built from excitatory and inhibitory (amacrine) pathways, which explains classic illusions of contrast. These fields also adapt within seconds to ambient light. A recent discovery from Marcus Meister's lab at Harvard shows that the retina already adapts to shapes. After seeing vertical bars, its response fades and then jumps when horizontal bars appear, so it signals novelty. Da Silveira contrasts two mechanisms that both explain this in words: a Hebbian "learning" mechanism, akin to memory, and a faster "fatigue" mechanism that protects against over-activity. Only mathematical models can tell them apart, by predicting which transition the retina should treat as the bigger novelty. He concludes that theory distinguishes principles of computation from biological detail, much as nouvelle cuisine replaced recipes with principles, and tells experimenters what to measure.

In the discussion, Changeux explains his idea of "minimal consciousness" common to wakeful mammals, the newborn being already richer than the mouse, and defends a single global workspace in which the final decision is taken, without excluding alternative pathways. On masking experiments, he places conscious access at 50 to 200 milliseconds, with lateral inhibition blocking stimuli from the workspace. He stresses individual variability and epigenesis, recalling a "theorem of variability" in which the same input-output relation can be reached by different circuits. He argues that social scientists can no longer study collective phenomena such as changes in mentality without the constraints imposed by brains. Da Silveira and Changeux debate whether the prefrontal cortex is more "random" than visual areas. All agree that future neuroscience of higher functions will need mathematical treatment. Hakim recalls that, twenty-five years earlier, Changeux had called his question about consciousness unscientific.

Excerpts