Answers are grounded in this paper and cite their sources.
The hippocampus is a central cognitive hub whose dysfunction contributes to most neurodegenerative dementias. Despite extensive knowledge of its anatomy, physiology and circuitry, a unifying account that links hippocampal dynamics to cognition and clinical syndromes remains lacking. Drawing on systems neuroscience and recent advances in generative artificial intelligence, we conceptualize the hippocampus as a latent diffusion engine that compresses sensory, interoceptive and cognitive inputs into low-dimensional representations and regenerates percepts, episodic memories and imagined scenes for cortical integration. Travelling waves of cortico-hippocampal oscillations, aligned with large-scale functional gradients, schedule and structure this generative process and organize distributed neural activity into coherent, semantically grounded constructs. We formalize this proposal as stochastic latent oscillatory diffusion (SLOD), a biologically grounded framework that maps hippocampal-cortical computations onto specific anatomical substrates and dynamical processes. While we do not suggest that the hippocampus implements diffusion algorithms per se, we show that its computations can be decomposed into homologous operations including compression, stochastic exploration, conditioning and reconstruction, whose formal structure closely parallels that of latent diffusion models. Finally, we demonstrate how major neurodegenerative dementias, including the Alzheimer's disease spectrum, dementia with Lewy bodies and frontotemporal dementia, can be understood as selective breakdowns of distinct components of this generative architecture.
No takes yet. Share an insight, caveat, or question.
Jones et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: