Purdue researchers: Machines learn like our brain

Purdue researchers: Machines learn like our brain

New York, Feb. 10, 2022

Purdue University researchers have unveiled a process that could form the basis for future artificially intelligent computers that learn like a human brain. The device consists of circuit components that can be reconfigured using electrical pulses. It could pave the way for neural networks in hardware that rewire themselves as they learn new information.

The researchers made the device from a material called perovskite nickelate.

Applying electrical pulses to the device caused it to shift hydrogen ion concentrations and create states that resemble human brain functions.

For example, when more hydrogen is near the center, it behaves like a neuron, while fewer hydrogen ions lead to a synapse. Simulations have shown that the device forms a structure for an artificial neural network.

The device could make artificial intelligence more mobile by embedding it directly into hardware instead of running it on software.

The team is now working to demonstrate the system on large-scale test chips.

This could form the basis for a future brain-inspired computer that can adapt and learn like a human.