When Solid-State Physics Inspires Artificial Intelligence: Meet Paul Saighi, LPS Alumnus

How can the brain revisit memories during sleep? And what if this mechanism could inspire new approaches to artificial intelligence?

Paul Saighi, a former PhD student at the Laboratoire de Physique des Solides (LPS, CNRS / Université Paris-Saclay) and the Institute of Biology of the École normale supérieure (IBENS), looks back on his doctoral research at the intersection of physics and computational neuroscience in an interview published by DIM AI4IDF.

Selected for the first DIM AI4IDF cohort in 2022, Paul Saighi investigated how memories are reactivated in artificial neural networks, drawing inspiration from brain activity during sleep.

When an artificial neural network learns new information, it can sometimes lose knowledge acquired previously. This phenomenon, known as “catastrophic forgetting”, is a major challenge for artificial intelligence. Paul Saighi’s research explores how the autonomous reactivation of memories, inspired by what happens in the brain during sleep, could help limit this forgetting and promote knowledge transfer between different parts of a network.

This research highlights the value of interdisciplinary approaches at the crossroads of physics, neuroscience and artificial intelligence. It also demonstrates the importance of fundamental research in exploring new ways to understand the brain and develop future AI models.

In the interview, Paul Saighi also shares his thoughts on the freedom to pursue research, interdisciplinary collaborations, the future of AI and his journey as an early-career researcher.

Read the full interview:

When Solid-State Physics Reinvents Artificial Intelligence – Interview with Paul Saighi, DIM AI4IDF