Neuronal firing patterns have significant spatiotemporal variability with no agreed-upon theoretical framework. Using a combined experimental and mode…

Classically, biological reproduction is explained as the building of a new organism from replicated genomic instructions. The corresponding theoretical model is von Neumann's self-reproducing machine, which relies on an invariant universal constructor that can build any machine from instructions. However, the reproductive incompatibility of species and the diversity of developing processes speak against the existence of a universal constructor. Without a universal constructor, the genome as representation of the organism is circularly defined: what the genome represents is specified by the developmental processes represented by the genome.<p>I propose to take invariant reproduction not as a premise, but as an emergent dynamical property.</p><p>Reproduction is seen as the iteration of a transform that maps one generation to the next, a transform shaped by the genome. Invariant reproduction then occurs when a reproductive sequence converges to a fixed point. A reproductive sequence may also diverge, converge to a cycle (multigenerational life cycle), or to one of several fixed points (non-genomic inheritance). When it does converge, it is necessarily to a stable point, implying that development is robust to perturbations. Finally, speciation can be understood as a process by which reproductive transforms become mutually incompatible, that is, the basins of attractions of the fixed points do not overlap any more. In this view, the genome is an inheritable constraint on development, not a representation of the organism. I suggest that this dynamical framework is a more coherent model of biological reproduction than von Neumann's computational framework.</p>

Traveling waves (TWs) are a fundamental mechanism of large-scale spatiotemporal coordination in the cortex, yet how their directionality originates from microscale circuit properties remains an open q...

Deep neural networks and brains both learn and share superficial similarities: processing nodes are likened to neurons and adjustable weights are likened to modifiable synapses. But can a unified theo...

As scientists, we want solid answers, but we also want to answer questions that matter. Yet, the brain’s complexity forces trade-offs between these desiderata, bringing about two distinct research app...

Long-term learning and memory in the primate brain rely on a series of hierarchically organized subsystems extending from early sensory neocortical areas to the hippocampus. The components differ in t...

How does genomic information unfold, to give rise to self-constructing living organisms with problem-solving capacities at all levels of organization? We review recent progress that unifies work in de...

Defining the minimal genetic requirements for cellular life remains a fundamental question in biology. Genomic exploration continually reveals novel microbial lineages, often exhibiting extreme genome...

Why language models succeeded where video models failed, and what that teaches us about AI

Computation in recurrent networks of neurons has been hypothesized to occur at the level of low-dimensional latent dynamics, both in artificial systems and in the brain. This hypothesis seems at odds ...

Proceedings of the National Academy of Sciences (PNAS), a peer reviewed journal of the National Academy of Sciences (NAS) - an authoritative source of high-impact, original research that broadly spans...

Mental maps of environmental structure enable the flexibility that defines intelligent behavior. We describe a striking internal representation in rat prefrontal cortex that exhibits hallmarks of a wo...

The inevitable emergence of good solutions?

Conscious artificial intelligence and biological naturalism

Sensory Horizons and the Functions of Conscious Vision

This article introduces and defends the “pathological complexity thesis” as a hypothesis about the evolutionary origins of minimal consciousness, or sentience, that connects the study of animal consci...

Recent work proposes using world models to generate controlled virtual environments in which AI agents can be tested before deployment to ensure their reliability and safety. However, accurate world m...

The renormalization group (RG) is a theoretical framework to transform systems across scales and identify critical points of phase transitions. In recent years, efforts have extended RG to complex net...

Join us to explore neuro-foundation models. March 31-April 1, 2025 in Mont Tremblant, Canada.

We seek to understand if there is interest in being able to attend NeurIPS in Europe, i.e. without travelling to San Diego, US. In the following, assume that it is possible to present accepted papers ...

Training deep neural networks typically relies on backpropagating high dimensional error signals a computationally intensive process with little evidence supporting its implementation in the brain. Ho...

My conversation with neuroscientist Paul Cisek

Implications draw on the history of transformative information systems from the past

A long-standing debate in neuroscience concerns whether individual neurons are organized into functionally distinct populations that encode information differently ("categorical" representations) and the implications for neural computation. Here, we systematically analyzed how cortical neurons encode cognitive, sensory, and movement variables across 43 cortical regions during a complex task (14,000+ units from the International Brain Laboratory public Brainwide Map data set) and studied how these properties change across the sensory-cognitive cortical hierarchy. We found that the structure of the neural code was scale-dependent: on a whole-cortex scale, neural selectivity was categorical and organized across regions in a way that reflected their anatomical connectivity. However, within individual regions, categorical representations were rare and limited to primary sensory areas. Remarkably, the degree of categorical clustering of neural selectivity was inversely correlated to the dime

How do Large Language Models Make Sense of Their “Worlds”?

YouTube video by Weinan Sun

Translaminar spacetime patterns organized by feedback-dependent cortical circuits support sparse traveling waves.

A major challenge in neuroscience is reconciling idealized theoretical models with complex, heterogeneous experimental data. We address this challenge through the lens of continuous-attractor networks...

The extent to which structural modularity in neural networks ensures functional specialization remains unclear. Here the authors show that specialization can emerge in neural modules placed under reso...

Using concepts mostly familiar to fluid physicists, the authors manage to elegantly tackle the complex problem of describing large-scale brain activity patterns, providing a rich foundation for the fu...

YouTube video by Kempner Institute at Harvard University

A neocortical–entorhinal–hippocampal network model based on grid cell states recapitulates experimental results and reconciles the spatial, associative and episodic memory roles of the hippocampus.