20 Oct 2025
On October 20, ASI Global welcomed Professor Edvard Moser, 2014 Nobel Laureate in Physiology or Medicine and Professor of Neuroscience at the Norwegian University of Science and Technology (NTNU), for the October session of the LUMINAI Lecture Series entitled “The Brain’s Codes for Space”. Addressing an online audience of Chinese college students and young scholars, Professor Moser provided an engaging overview of how the brain constructs an internal map of the world.
Professor Moser opened the lecture by positioning spatial navigation as one of the brain’s most fundamental cognitive abilities. He traced the field’s history from early behavioral theories of “cognitive maps” to physiological discoveries that provided a concrete neural basis for these ideas. He highlighted three specific types of neurons that act as the pillars of this system: place cells in the hippocampus that fire at specific locations; head-direction cells in the presubiculum that track orientation and direction, and grid cells discovered by the Moser lab in 2005 in the medial entorhinal cortex that covers the entire space available to the animal, functioning like a coordinate system.
Discovery of Grid Cells
The lecture then turned to the core, with a focus on the structure and computation of grid cells. Professor Moser explained that grid cells are organized into modules, each encoding space at a different scale. When combined, these modules provide a multi-resolution representation, allowing for precise navigation across environments of varying sizes. Significantly, research using topological analysis has revealed that the activity patterns of grid cells map onto a toroidal (donut-shaped) geometry. This structure reflects the intrinsic relationships between neurons. Because this grid-like activity appears early in development and persists during sleep or in the absence of visual cues, it suggests that the spatial system is a built-in feature of the brain rather than a product of learning.
Spatial Map’s Intrinsic Origin
Another major theme of the lecture was the deep connection between spatial navigation and memory, as the same entorhinal-hippocampal circuits that map space play a central role in forming episodic memories. While the entorhinal cortex is characterized by a single coherent map, the hippocampus consists of many uncorrelated maps. The transition between these states is enabled by the independent translation of grid modules, which creates an enormous number of orthogonal activity combinations. This flexible coding mechanism allows the brain to distinguish between similar experiences, a feature crucial for storing and retrieving personal memories. Professor noted that this relationship has important clinical implications: because the entorhinal cortex is among the first regions affected in Alzheimer’s disease, early spatial disorientation in patients may directly reflect damage to these fundamental circuits.
From Grid Cells to Memory
Toward the end of the lecture, Professor Moser reflected on connections and differences between biological intelligence and artificial intelligence. While both systems use distributed, modular representations, the brain achieves these functions with remarkable energy efficiency within strict biological constraints. He emphasized mutual benefit: principles such as multi-scale encoding in grid cells could inspire new approaches to AI navigation and robotics. At the same time, modern AI tools, particularly in pattern recognition, are becoming indispensable for analyzing large-scale neural data. Professor Moser emphasized that progress in both fields increasingly depends on cross-fertilization.
Comparison of Brains and Computers
During the Q&A session, students asked how the brain’s navigation differs from GPS technology. Professor Moser explained that while both systems use movement signals to track position, the brain operates with extraordinary energy efficiency. He also discussed how grid cell research could help detect early signs of Alzheimer’s disease and whether biological navigation principles could improve robotic systems, noting that optimal solutions may differ due to distinct biological and engineering constraints.