Researchers develop transistor with brain-like functionality

(CN) — Brain-like computers may not be purely science fiction for long, as researchers have created a new synaptic transistor capable of higher-level thinking.

That’s according to a new study published in the journal Nature on Wednesday, which reveals how researchers from Northwestern University, Boston College and the Massachusetts Institute of Technology designed the memory resistor or “memristor” to surpass simple machine-learning tasks to perform much like the human brain.

“The brain has a fundamentally different architecture than a digital computer,” said co-author Mark Hersam, from Northwestern University, in a statement. “In a digital computer, data move back and forth between a microprocessor and memory, which consumes a lot of energy and creates a bottleneck when attempting to perform multiple tasks at the same time.”

When it comes to brains, however, Hersam explains that memory and information processing are colocated and fully integrated, enabling higher orders of magnitude with greater energy efficiency.

“Our synaptic transistor similarly achieves concurrent memory and information processing functionality to more faithfully mimic the brain,” Hersam said.

Unlike previous studies that have produced similar brain-like devices, the memristor can perform outside cryogenic temperatures while demonstrating stability with faster, more energy efficient functioning at room temperature. The device also retains stored information without a power source, making it ideal for real world applications.

To create such a device, the researchers had to think outside the paradigm in electronics that involves building everything out of transistors while using the same silicon architecture.

“Significant progress has been made by simply packing more and more transistors into integrated circuits,” Hersam said. “You cannot deny the success of that strategy, but it comes at the cost of high-power consumption, especially in the current era of big data where digital computing is on track to overwhelm the grid. We have to rethink computing hardware, especially for AI and machine-learning tasks.”

In rethinking their own hardware, the researchers incorporated the physics of moiré patterns, a geometrical design created when layering two patterns on top of one another. They combined bilayer graphene and hexagonal boron nitride before carefully stacking and twisting the materials for neuromorphic functionality.

“With twist as a new design parameter, the number of permutations is vast,” Hersam said. “Graphene and hexagonal boron nitride are very similar structurally but just different enough that you get exceptionally strong moiré effects.”

While testing the transistor, the researchers trained it to recognize similar numeric patterns, such as “000” from “111” and “101.”

“If AI is meant to mimic human thought, one of the lowest-level tasks would be to classify data, which is simply sorting into bins,” Hersam said. “Our goal is to advance AI technology in the direction of higher-level thinking. Real-world conditions are often more complicated than current AI algorithms can handle, so we tested our new devices under more complicated conditions to verify their advanced capabilities.”

The test was a success, Hersam said, explaining that by training the device to identify three zeros, it could then recognize that “111” is more similar to “000” than “101.”

“000 and 111 are not exactly the same, but both are three digits in a row,” Hersam said. “Recognizing that similarity is a higher-level form of cognition known as associative learning.”

Even with incomplete patterns, the researchers say the transistor demonstrated associative learning.

“Current AI can be easy to confuse, which can cause major problems in certain contexts,” Hersam said. “Imagine if you are using a self-driving vehicle, and the weather conditions deteriorate. The vehicle might not be able to interpret the more complicated sensor data as well as a human driver could. But even when we gave our transistor imperfect input, it could still identify the correct response.”

Follow @alannamayhampdx