Researchers led by Indian American scientist are taking inspiration from the human brain to develop computing architecture that can support the growing energy demands of artificial intelligence.

“There’s nothing in the world that’s as efficient as our brain — it’s evolved to maximize the storage and processing of information and minimize energy usage,” says Sambandamurthy Ganapathy,  professor in the University at Buffalo Department of Physics and associate dean for research in the UB College of Arts and Sciences.

“While the brain is far too complex to actually recreate, we can mimic how it stores and processes information to create more energy-efficient computers, and thus, more energy-efficient AI.”

While an AI model is estimated to take over 6,000 joules of energy to generate a single text response, human brain needs just 20 joules every second to keep one alive and cognitive.

READ: 8 Indian Americans among 10 3M Young Scientist Challenge finalists (June 26, 2025)

This brain-inspired approach is known as neuromorphic computing. Its origins go back to the 1980s but it has taken on more relevance in recent years as computing tasks have become more energy intensive and complex, especially tasks that require AI, according to a university release.

While neuromorphic computing can relate to both brain-inspired hardware and software, Ganapathy’s team is focused on hardware.

Their research, funded by the National Science Foundation, is a blend of quantum science and engineering that involves probing the unique electrical properties of materials that can be used to build neuromorphic computer chips.

The team’s goal is to ultimately develop chips and devices that are not only more energy efficient, but also just better at completing tasks — perhaps even in a more human-like way.

“The computers of today were built for simple and repetitive tasks, but with the rise of AI, we don’t want to just solve simple problems anymore,” Ganapathy says. “We want computers to solve complex problems, like human beings do every day. Neuromorphic computing may provide the structure to allow computers to do this.”

“Neuromorphic computing simply aims to move beyond the binary framework and closer to the far more complex system given to us by nature,” says Nitin Kumar, a graduate student in Ganapathy’s lab.

READ: Four Indian Americans named 2025 Guggenheim fellows (April 18, 2025)

One of the ways the brain is more complex — and energy efficient — than a computer is that information is stored and processed in the same place.

“It’s not as if the left side of the brain holds all the memories and the right is where all learning happens,” Ganapathy says. “It’s intertwined.”

Information storage and processing are separated in traditional computers, and thus, a lot of energy is used simply transporting data along tiny circuits between its memory unit and its processing unit. This can become even more energy-intensive when the computing architecture is supporting an AI model.

“Of course, the question then becomes how close we can place memory and processing together within a computer chip,” Ganapathy says. “This is known as in-memory computing and it’s a major advantage of neuromorphic computing.”

Memory and processing are intertwined in the brain thanks to an intricate system of neurons. So Ganapathy’s team is developing artificial neurons and synapses designed to mimic their biological counterparts’ electrical signaling of information.

READ: Har Gobind Khorana: Indian American scientist who cracked the genetic code (January 10, 2022)

“We essentially want to recreate those rhythmic and synchronized electrical oscillations you may see in a brain scan,” Kumar says. “To do this, we need to create our neurons and synapses out of advanced materials whose electrical conductivity can controllably be switched on and off with precision.”

“Our next goal,” Ganapathy adds, “is to synchronize the oscillations of multiple devices to construct an oscillatory neural network capable of emulating complex brain functions such as pattern recognition, motor control and other rhythmic behaviors.”

Ganapathy stresses that neuromorphic computers mimic the brain on a purely phenomenological level. Neuromorphic computing aims to recreate the brain’s functional behaviors and benefits — not consciousness.

However, it’s possible that neuromorphic computers will solve problems less like computers and more like human beings.

Researchers think this could be especially helpful in applications like self-driving cars, where AI does well in most road situations but still underperforms humans when it comes to more complex scenarios with no easy solution.

“Neuromorphic chips may not be in your smartphone anytime soon, but I do think we will see them in highly specific applications, like self-driving cars. Perhaps even one chip to respond to the road and another to find the best possible route,” Ganapathy says.