Tech talk
Efficient neural coding under resource constraints
Tech Talk with Leo D’Amato
🗓️ When: July 23, 2025, at 3:00 PM CEST
🎤 Speaker: Leo D’Amato, Research Fellow at the Institute for Cognitive Science and Technologies of the National Research Council of Italy (ISTC-CNR).
📍 Where: Online on Google Meet. Register now!
Why do intelligent systems—whether biological or artificial—develop distorted internal representations of the world? What is the impact of limited resources on complex cognitive abilities like number sense?
Join us for an engaging tech talk with Leo D’Amato, Research Fellow at ISTC-CNR and Ph.D. in Artificial Intelligence, who will explore how inherent resource limitations don’t just hinder intelligence—they shape it.
Using insights from Rate-Distortion Theory (RDT) and hands-on experimentation with β-Variational Autoencoders (β-VAEs), Leo will demonstrate:
- How resource constraints lead to specific geometric distortions in the latent representations of generative models.
- Why key phenomena like prototypisation, specialisation, and orthogonalisation emerge as principled adaptations to RDT constraints.
- How number sense—a fundamental cognitive ability—can arise spontaneously in AI models without supervision.
This talk bridges cognitive science, artificial intelligence, and information theory, revealing how compressed, efficient representations can still unlock deep intelligence in machines and, perhaps, in ourselves.
About the speaker
-
Leo D'Amato
Research Fellow at ISTC-CNR and Ph.D. in Artificial Intelligence
Leo D’Amato is a Research Fellow at the Institute for Cognitive Science and Technologies of the National Research Council of Italy (ISTC-CNR). His work lies at the intersection of cognitive science and artificial intelligence, focusing on computational models for complex cognitive abilities such as number sense, episodic memory, and spatial navigation.
He holds a Ph.D. in Artificial Intelligence from the Polytechnic University of Turin and has a background in mathematics. Before starting his Ph.D., he gained industry experience. He has recently contributed to several applied research projects, including acoustic emission analysis for structural health monitoring of buildings and railway traffic management based on self-organising multi-agent systems.