Generative AI breakthroughs in a recent UCL study, funded by Wellcome, explore how memories shape learning and imagination. The focus is on the interplay between the hippocampus and neocortex, crucial for memory and planning.
Replaying memories aids pattern recognition for survival predictions. Lead author Eleanor Spens, a UCL PhD student, emphasizes the convergence of AI generative networks with human cognition.
Researchers exposed the model to 10,000 scenes, observing rapid encoding by the hippocampal network. This information trained the neocortical generative neural network, efficiently recreating scenes and generating new ones by capturing conceptual meaning.
Professor Neil Burgess, the senior author, details how memories reconstruct with a focus on meaning rather than veridical records. The model explains the acquisition of conceptual knowledge by the neocortex, collaboration with the hippocampus in “re-experiencing” events, and memory bias understanding.
Source NeuroScienceNews
