A Model for Computers with Human-Like Memories
on 07 October 2016
Thus far, we've been able to replicate both a cat and monkey-sized artificial brain. The human brain, however, has proven more elusive—though some scientists say it can be done by 2019. This blockade is energy. According to the Atlantic, artificial brains require so much energy as they have the capability to be infinitely precise and draw upon epic amounts of data. Take, for example, how human's recognize another human's face—and multiply the accuracy based on a lifetime's worth of memories. Theoretical neuroscientists Stefano Fusi and Marcus Benna offered up a solution in a new paper published in Nature Neuroscience: move toward stability. In humans, synapses in the brain form new memories while protecting older ones at the same time. There is a metaplasticity that allows us to absorb more into a chaotic mind at a young age, while later in life memories and knowledge are, for lack of better word, "hardened." These variables and the model based off of it might save energy and allow for a breakthrough in developing artificial human-like brains.