Spin-based memory cells can be trained to learn like a brain synapse
Researchers from France's Universite Paris-Sud and the CEA institute discovered that the probabilistic nature of Spin-Torque MRAM (STT-MRAM) devices can be used to create synapses-like neural system - effectively to create low-power devices that mimic the human's brain method of operation.
MRAM cells (or MTJs) store data using electrons magnetic spin, which is a stochastic switching type of device - that needs to apply a current for a long-enough time to make sure the information changed as you wish. In this new suggestion, you take advantage of the stochastic switching and apply current for a short time - which can be used to make the device learn progressively. This effectively means that MTJs can be "trained" to learn new information.