Berlin Learning &
Intelligent Systems Society

This summer semester we are hosting a BLISSful AI reading group. Various BLISS members will present papers related to their research.

Location: Merantix AI Campus (Max-Urich-Straße 3, 13355 Berlin)

Time: Every Monday, 6:45pm

Abstract

A central goal of systems neuroscience is to understand how populations of neurons represent and transform the natural world. As neural datasets grow in scale and complexity, manual hypothesis generation and neuron-by-neuron characterization are becoming increasingly limiting. Neural digital twins — deep learning models trained to predict the responses of biological neurons — offer a complementary approach: distilling large-scale recordings into experimentally tractable models that can be queried, perturbed, and analyzed in silico. Using this framework, we characterize neurons in macaque and mouse visual cortex, revealing interpretable feature selectivity and previously hidden properties of the neural code. More broadly, these models enable a paradigm of virtual neuroscience, where learned models of neural systems become platforms for automated scientific discovery: searching stimulus space, mapping selectivity, testing representational hypotheses, and generating new candidate principles of neural computation. By enabling virtual experiments that would be difficult or impossible in vivo, neural digital twins open a path toward a new regime of scalable, computationally driven discovery in neuroscience.

About

Sophia Sanborn is an Assistant Professor in the Department of Ophthalmology, with a courtesy appointment in Neurology, at Stanford University. She is Co-Director of the Enigma Project, a large-scale neuroscience initiative aimed at understanding the neural code, and Co-Founder of Metamorphic, an AI company building neural digital twins and brain-aligned foundation models. Her research sits at the intersection of neuroscience, artificial intelligence, and mathematics, leveraging large-scale neural data and computational models to reveal principles of neural coding in biological and artificial systems.