Just like a Black Mirror episode, Samsung and Harvard have published research suggesting it’s possible to make a brain-inspired memory chip.
The paper published in Nature Electronics proposed a method that would essentially “copy and paste” a brain’s neuron wiring map to 3D neuromorphic chips.
By using a nanoelectrode array (an electrical conductor that carries electricity), they believe the brains neuronal connection map could be copied.
This nanoelectrode array could record electrical signals produced by a large number of neurons in the brain. This would then tell the brain’s neuronal map where neurons connect with one another and how strong the connections are, the researchers claimed.
You could then essentially copy the data and ‘paste’ it to a 3D network of solid-state memory.
Samsung Electronics Puts Forward a Vision To ‘Copy and Paste’ the Brain on Neuromorphic Chipshttps://t.co/pjEk6yZWNN
— Samsung Electronics (@Samsung) September 25, 2021
The ability to do this could serve as a ‘shortcut’ to creating artificial intelligence systems that behave like real brains, including the ability to learn new concepts, adapt to changing conditions and learn simple things.
The paper also suggests that directly downloading the map onto a memory chip could be a faster way of pasting the neuronal map.
“The vision we present is highly ambitious,” said Donhee Ham of Samsung Advanced Institute of Technology, “But working toward such a heroic goal will push the boundaries of machine intelligence, neuroscience, and semiconductor technology.”
This has the potential to create fully autonomous machines with true cognition, according to the researchers.
The concept of robots with full cognition is both impressive and slightly concerning, particularly when considering the extremities depicted in countless Hollywood movies, books and TV shows.
But then again, those are just fictional, right?
There is no way this could possibly become any sort of huge disaster.
— Geoff Monty (@MontyGeoff) September 27, 2021
While the concept is exciting, the complexity of performing all of this is the biggest issue.
The human brain has roughly 100 billion neurons with a thousand times more synaptic links. This means the chip would need about 100 trillion memory units, which is over 100 terabytes or 100,000 gigabytes of data.
This alone is a huge challenge, regardless of the complex coding needed to make the virtual brain work.
While Samsung and Harvard have opened a door to realistic AI, it seems like it will be some time before their concept becomes reality.