Final Report Summary - M4 (Memory Mechanisms in Man and Machine)
We have probably all had the experience of recognizing something we have not seen or heard for a very long time. In older people, such memories can clearly survive for decades, even with no chance to reactivate those memories to keep them alive. How does the brain keep such memories intact for such long periods? The M4 project (Memory Mechanisms in Man and Machine) has looked at the idea that these memories could involve the formation of highly selective neurons or “Grandmother Cells” that only respond to very specific stimuli, and that these neurons can remain silent for long periods of time, waiting for a previously learned stimulus to reoccur. We argue that such selective neurons can be formed with just a few presentations of a newly encountered stimulus, and indeed, this ability to detect repeating patterns may be a key to understanding natural intelligence. Furthermore, we have developed learning rules that allow networks of artificial neurons to do the same trick, opening up the possibility of developing new types of electronic devices that can mimic the remarkable learning capacities of the human brain.