University College London (UCL) researchers are pioneering the use of generative artificial intelligence (AI) to delve into the complexities of human imagination, learning, and memory recall.
AI Mirrors Cognitive Brain Functions
Published in "Nature Human Behaviour," the UCL study harnesses a generative AI neural network to mimic the brain's neural pathways, particularly focusing on the hippocampus and neocortex. The research examines how these brain regions collaborate for memory recall, imaginative thought, and planning.
Eleanor Spens, a Ph.D. candidate at UCL's Institute of Cognitive Neuroscience, points to the sophistication of AI generative networks. These networks are adept at drawing from past experiences, enabling them to reconstruct specific memories and conjure up novel scenarios. This process is akin to "imaginative remembering," where the brain pieces together stored information and probable outcomes to form a coherent memory.
Predictive Function Essential for Survival
The human brain's predictive capability is critical for survival, helping individuals avoid threats or secure resources. According to AI models, when the brain processes memories during periods of rest, it identifies patterns from past events to enhance future prediction accuracy.
Hippocampus' Role in Encoding and Recreating Scenes
The study exposed an AI model to 10,000 simple scenes, with its hippocampal network encoding the scenes for later replay. This replay trained the neocortical network, teaching it to interpret visual information and recreate these scenes as activity patterns in numerous output neurons.
The neocortical network's learning of conceptual scene representations not only allows for the recollection of past scenes but also the invention of new ones. The hippocampus prioritizes encoding the essence of scenes rather than their exhaustive detail, highlighting the brain's focus on distinctive features. The model provides insight into how the neocortex accumulates conceptual knowledge and collaborates with the hippocampus in reconstructing mental events.
Professor Neil Burgess, a senior author from UCL, elaborates on the reconstructive nature of memories. He suggests that memories are not precise replicas of past events but are interpretations enriched with meaning and specific details. This reconstruction process reveals the inherent biases in memory recall and the impact of personal experiences on how memories are formed and remembered.
Read More...