The Key to Lifelong Learning in the Brain Is Now Available as Hardware for AI

The Key to Lifelong Learning in the Brain Is Now Available as Hardware for AI

The human brain adjusts as it picks up new information. However, as soon as artificial intelligence picks up a new skill, it frequently forgets previous knowledge.

A paper published in Science this week demonstrates a way that computer chips could dynamically rewire themselves to take in new data like the brain does, helping AI to keep learning over time.

As businesses use more and more data to improve how AI recognizes images, learns languages, and performs other complex tasks, this method could help AI continue to learn as time goes on.

“The brains of living beings can continuously learn throughout their lifespan. We have now created an artificial platform for machines to learn throughout their lifespan,” said Shriram Ramanathan, a professor in Purdue University’s School of Materials Engineering who specializes in discovering how materials could mimic the brain to improve computing.

The circuits on a computer chip remain the same, unlike the brain, which continuously creates new connections between neurons to facilitate learning. The circuit that was initially created for a machine in a factory is identical to the circuit that a machine has been utilizing for years.

Making AI more portable presents a challenge since it would require autonomous vehicles or robots in space to make judgments on their own in remote settings. These machines would be more productive if AI could be built directly into hardware as opposed to only running on software, as it currently does.

In this study, Ramanathan and his team developed new technology that can be quickly and easily reprogrammed using electrical pulses. Ramanathan thinks that the device’s versatility would enable it to perform every task required to create a brain-inspired computer.

“If we want to build a computer or a machine that is inspired by the brain, then correspondingly, we want to have the ability to continuously program, reprogram and change the chip,” Ramanathan said.

The brains of living beings can continuously learn throughout their lifespan. We have now created an artificial platform for machines to learn throughout their lifespan.

Professor Shriram Ramanathan

Toward building a brain in chip form

A little rectangular gadget known as perovskite nickelate, which is extremely sensitive to hydrogen, serves as the hardware. In a matter of nanoseconds, the gadget can shuffle a concentration of hydrogen ions by applying electrical pulses of varying voltages, producing states that the researchers discovered could be linked to corresponding brain activities.

For instance, the gadget can function as a neuron, or a single nerve cell, when it has more hydrogen towards its center. The structure acts as a synapse, a connection between neurons, which is what the brain uses to retain memories in intricate neuronal circuits, because there is less hydrogen at that site.

Collaborations between the Santa Clara University and Portland State University and the Purdue University team have demonstrated that the internal physics of this device create a dynamic structure for an artificial neural network that is able to recognize electrocardiogram patterns and digits more quickly than static networks.

This neural network makes use of “reservoir computing,” which clarifies how various brain regions connect with one another and exchange information.

In this study, researchers from The Pennsylvania State University also showed how a dynamic network may “choose and choose” which circuits are best suited for solving particular issues as they arise.

Ramanathan thinks the semiconductor industry can quickly embrace this technique because the team was able to construct the gadget using common manufacturing methods that are compatible with semiconductors and run it at ambient temperature.

“We demonstrated that this device is very robust,” said Michael Park, a Purdue Ph.D. student in materials engineering. “After programming the device over a million cycles, the reconfiguration of all functions is remarkably reproducible.”

On large-scale test chips that would be used to construct a computer inspired by the human brain, the researchers are striving to illustrate these principles.

Experiments at Purdue were conducted at the FLEX Lab and Birck Nanotechnology Center of Purdue’s Discovery Park. Measurements of the device’s characteristics were made by the team’s partners at Argonne National Laboratory, the University of Illinois, Brookhaven National Laboratory, and the University of Georgia.

The research was supported by the U.S. Department of Energy Office of Science, the Air Force Office of Scientific Research and the National Science Foundation.