Making computers more brain-like: Stanford researchers create first artificial synapse

Centuries of evolution have made the human brain the most power efficient machine in existence.  Despite the fact that the human brain only generates enough electricity to power a LED light, it can move information at a speed of 268 mph and has a storage space that is estimated to be close to 2.5 petabytes (2.5 million gigabytes). It is no wonder that scientists are working to create computers that imitate the brain’s efficiency.  

How does the brain learn? 

In a pioneering step towards creating a brain-like computer, a team of Stanford University researchers developed the first artificial synapse in 2017. A synapse is the area between two neurons. Whenever we learn something new, electrical signals travel down one neuron and cause the release of chemicals called neurotransmitters. Neurotransmitters transverse the synapse and binds to receptors on the second neuron essentially relaying the electrical signal. The most energy is expended the first time a synapse is traversed as each consequential connection requires less energy. This is how the brain efficiently learns and remembers everything that is learned.  

How does the artificial synapse work? 

The artificial synapse, which was built based on the design of a battery and made with organic materials, replicates the brain's learning process. Researchers continuously discharged and recharged the artificial synapse to emulate the way neural connections are strengthened by learning. This allowed the scientists to predict the voltage required to excite the synapse to a particular electrical state. Once it reached that point, the artificial synapse remained at that state.   

This will allow future computers to recall your work without any additional actions. The kinds of tasks we expect our computing devices to do require computing that mimics the brain because using traditional computing to perform these tasks is becoming really power hungry. "We've demonstrated a device that's ideal for running these ... algorithms and that consumes ... less power," says A. Alec Talin, one of the senior authors of the project’s paper.