In an innovative leap at University College London (UCL), funded by the Wellcome Foundation, researchers are bridging the gap between artificial intelligence and human cognition using generative AI.
Led by Ph.D. student Eleanor Spens, the team employed generative AI models to simulate the intricate process of how our brains encode, store, and recall memories. This advancement is essential in understanding our capability to learn, relive past experiences, and construct new scenarios for imagination and planning.
The core of the study focused on the utilization of generative neural networks, which were designed to mimic the functions of the hippocampus and neocortex in the brain.
These areas are crucial for memory, imagination, and planning. By training the AI model with a series of events represented as simple scenes, the researchers gained insight into the formation, replay, and imaginative use of memories.
Eleanor Spens, lead author of the study and a Ph.D. student at UCL, said, “Recent advances in the generative networks used in AI show how information can be extracted from experience so that we can both recollect a specific experience and also flexibly imagine what new experiences might be like.”
A particularly fascinating aspect of the study was how it demonstrated the importance of memory replay. The AI model was exposed to 10,000 images of simple scenes, with the hippocampal network rapidly encoding each scene as it occurred. This continual replay of memories is believed to help the brain identify patterns from past experiences, essential for making survival-related predictions.
The implications of this research are vast, especially in the field of cognitive science. The AI’s neocortical network learned to process the scenes into efficient “conceptual” representations, allowing for the reconstruction of old scenes and the generation of entirely new ones.
This breakthrough provides a clearer understanding of how the neocortex acquires knowledge and, alongside the hippocampus, enables us to reconstruct and relive past events in our minds.
This subject has long been a focal point of debate, and invariably, opinions diverge regarding the capability of generative AI to emulate human thought and functionality.
Some of the human brain processes are likely roughly similar to how current generative AI models work, but there are still big differences.
— Stefan Andersson (@Stefan3334) December 22, 2023
This is why I honestly hate the recent ‘AI’ trend, because none of it is actually intelligent, they’re just more advanced adaptive algorithms that still only ‘learn’ based off whatever nonsense you pump into it. That’s not intelligent, that’s just good coding.
— Rampaging Redbeard (@RedRaptorJockey) January 10, 2024
Extending beyond the laboratory, the study also sheds light on the broader applications of generative AI in understanding human memory and imagination.
Generative AI’s ability to produce text, images, and other media forms is now channeled into visualizing the relationship between visual archives and human memory.
This novel approach is critical for understanding memory formation and recall, particularly in memory disorders such as Alzheimer’s disease.
The convergence of artificial intelligence with cognitive research is proving to be a ground-breaking domain for discovery. As generative AI continues to evolve, its potential to decode the complexities of memory and imagination is expanding.
For more AI news and insights, visit the news section of our website.
