Imperial College ICARL Seminars - Zafeirios Fountas (Huawei)

We are excited to continue our series with Zafeirios Fountas (Huawei) who will be talking about Episodic Memory, Consolidation, and Synergistic Processing: Convergent Principles in Biological and Scalable Artificial Intelligence.

The talk will be hybrid and you will be able to listen and ask questions both in person or remotely. After the talk there will be a Q&A and a catered networking session. Please note that the link to the session will be available at the end of the confirmation email, where it says "Additional information".

Abstract

A fundamental gap that remains between large language models and biological intelligence concerns the ability to learn continually from sparse experience whilst maintaining generalisation. I argue this gap stems from the absence of principled memory systems. This talk will present a unified framework connecting event-based episodic memory formation, consolidation, and information integration across biological and artificial systems. I will show that surprise-driven episodic segmentation triggers memory boundaries and produces event structures in LLMs that correlate with human perception, while predictive forgetting (outcome-conditioned compression of memories) provably optimises generalisation in high-capacity systems. These principles yield scalable AI systems that exceed current performance in generalisation, computational efficiency, and memory capabilities. Beyond these advances, I will show that LLMs spontaneously develop synergistic information processing cores mirroring human cortical organisation, suggesting that memory-driven information integration may be a fundamental signature of intelligence, with implications for both scalable AI and theories of consciousness.

About the speaker

Zafeirios Fountas is a Principal Research Scientist and Group Leader at Huawei's Noah's Ark Lab in London, leading research on brain-inspired AI and memory systems. He holds a PhD in Computational Neuroscience from Imperial College London, has held research positions at Imperial and UCL, and spent five years as lead AI scientist at the award-winning robotics startup Emotech. His work spans episodic memory, predictive coding, active inference, and large language models, with publications at ICLR, NeurIPS, ICML, and Nature Communications.

Sponsors

This event is co-hosted by DocSoc and sponsored by Iconic.

Event Venue & Nearby Stays

room 144, 180 Queen's Gate, 180 Queen's Gate, London, United Kingdom

Concerts, fests, parties, meetups - all the happenings, one place.

Get App