What can electronic brains dream about?
It only seems fitting that IBM (in conjunction with 5 other major universities and DARPA), the company that brought us Jeopardy champ, artificial intelligence computer, Watson, would be who brings us something straight out of science fiction novels.
The NeuroSynaptic Chip
The first of its kind to process in the same fashion as the brain does; a “neurosynaptic core” with integrated memory (replicated synapses), computation (replicated neurons) and communication (replicated axons), all in one chip. It’s like taking computer computation from 2-Dimensions into full 3-Dimensional space. This attempts to overcome what is know as the “von Neumann paradigm”, the current way our present day’s computer architecture is ruled by. Von Neumann introduced the architecture of the processor and the memory being two separate pieces of hardware in the 1940s. By integrating the memory into the same hardware as the processor you begin to see context dependent processes in an energy efficient manor PLUS eliminating the bottle neck of the Bus…just like a brain.
And that seems to be one of the driving forces behind IBMs new chip, is the energy efficiency, and rightly so. They can already slap together simulate synapses and firing neuron simulations into a super computer. But even as powerful and helpful as a super computer can be, its downfall is its size, its upkeep, its administration and the amount of energy it takes to drive.
But as Dr. Dharmendra Modha, head of the SyNAPSE project (Systems of Neuromorphic Adaptive Plastic Scalable Electronics), says, the brain is not a neural network, it’s a synaptic network, if you look at the junctions between neurons (called synapses’) you would see that they outnumber the number of neurons by a factor of 10,000” What DARPA has asked the team was to demonstrate at a nano scale, low power material that captures the function of the synapse. A calculate size of 1 picojoule fitting 10,000,000,000 single synapses’ inside the space of 1 square centimeter
This interview which explains more that Dr. Modha did with Fast Company is fantastic and highly suggested.
That alone was 3 years ago.
Fast forward to today they have 2 working prototype designs, both at 45 total nanometers a piece and currently contain 256 neurons. One core contains 262,144 programmable synapses and the other contains 65,536 learning synapses. 3 years ago, it was an idea, today, it is quite the reality. What do you think 20 – 50 years will look like from now?
Sources IBM Press Release 
Images: IBM Research – Almaden