Learning dynamics & effective theories
What governs learning and the development of useful representations in neural networks, and how can effective theories and ideas from physics describe that process?
Institute for Fundamental AI research
1609 Institute
Investigating the mathematical, physical and statistical principles underlying intelligence and learning.
Website (coming soon)
The initiative
Understanding
intelligence
An institute for fundamental AI research, bringing a physicist’s perspective to questions about how intelligent systems learn, represent information and generalize.
The 1609 Institute connects methods from physics and cosmology with the study of artificial intelligence, focusing on how complex systems encode information and how models can be tested against data.
Research directions
The foundations
of learning
Information, inference & hybrid representations
How is information encoded, transformed and retained by a model, and how can complementary approaches to representation learning be combined to capture structure in data?
Learned structure, generalization & scientific models
How do architecture and internal organization shape a learned system’s behavior, and which principles explain its ability to generalize beyond its training examples?
Research contact
Start a conversation
For inquiries about the 1609 Institute and research collaborations.
alvise@1609institute.com