Home Career Using electricity to find materials that can ‘learn’ – ScienceDaily

Using electricity to find materials that can ‘learn’ – ScienceDaily


Scientists used the Advanced Photon Source to observe how inanimate material mimics learning-related behaviors, paving the way for better artificial intelligence.

Scientists looking to build the next generation of supercomputers are looking to the most complex and energy-efficient computer ever created: the human brain.

In some of their first forays into building brain-inspired computers, researchers are looking at a variety of non-biological materials whose properties can be tuned to show evidence of learning-like behavior. These materials can form the basis of hardware that can be combined with new software algorithms to create more powerful, useful and energy-efficient artificial intelligence (AI).

In a new study led by scientists at Purdue University, researchers exposed oxygen-deficient nickel oxide to brief electrical pulses and induced two distinct electrical responses similar to learning. The result is an all-electric drive system that exhibits these learning behaviors, said Rutgers University professor Sriram Ramanathan. (At the time of this work, Ramanathan was a professor at Purdue University.) The research team used the resources of the Advanced Photon Source (APS), a US Department of Energy (DOE) Office of Science user facility at DOE’s Argonne National Laboratory.

The first response, habituation, occurs when the material “gets used” to being lightly pinched. The scientists noticed that although the resistance of the material increases after the initial shock, it quickly becomes accustomed to the electrical stimulus. “The habituation is similar to what happens when you live near an airport,” said Fanny Radolakis, a physicist and beamline scientist at APS. “The day you move, you think ‘what a racket’, but in the end you hardly notice.”

Another response that the material shows, sensitization, occurs when a higher dose of electricity is administered. “With a larger stimulus, the response of the material increases rather than decreases over time,” Radolakis said. “It’s like watching a scary movie and then someone goes ‘boo!’ because of the angle – you can see that he really jumps.”

“Virtually all living organisms exhibit these two characteristics,” Ramanathan said. “They really are a fundamental aspect of intelligence.”

These two behaviors are controlled by quantum interactions between electrons, which cannot be described by classical physics, and which help create the basis for the phase transition in the material. “An example of a phase transition is when a liquid becomes a solid,” Radolakis said. “The material we’re looking at is right on the edge, and the competing interactions that happen at the electronic level can easily be tipped one way or the other by small stimuli.”

Having a system that can be controlled entirely by electrical signals is essential for brain-inspired computing applications, Ramanathan said. “Being able to manipulate materials in this way will allow the hardware to take over some of the responsibility for the intelligence,” he explained. “Using quantum properties to embed intelligence into hardware represents a key step toward energy-efficient computing.”

The distinction between habituation and sensitization may help scientists overcome a problem in AI design called the stability-plasticity dilemma. Artificial intelligence algorithms can often be, on the one hand, too reluctant to adapt to new information. But on the other hand, when they do, they can often forget some of what they’ve already learned. By creating habitable material, scientists can teach it to ignore or forget unnecessary information and thus achieve additional stability, while sensitization can teach it to remember and incorporate new information, providing plasticity.

“Artificial intelligence often has a hard time learning and retaining new information without overwriting information that has already been stored,” Radolakis said. “Too much stability prevents AI from learning, but too much plasticity can lead to catastrophic forgetting.”

One of the main advantages of the new research was the small size of the nickel oxide device. “This type of learning has not been done before in today’s generation of electronics without a lot of transistors,” Radolakis said. “This single-coupled system is the smallest system to date that exhibits these properties, which is of great importance for the possible development of neuromorphic circuits.”

To reveal the atomic-scale dynamics responsible for the habituation and sensitization behavior, Argonne’s Radolakis and Hua Zhou used X-ray absorption spectroscopy at the 29-ID-D and 33-ID-D APS beamlines.

A paper based on the research was published in the September 19 issue Advanced intelligent systems.

The research was funded by the Department of Energy’s Office of Science (Office of Basic Energy Sciences), the Army Research Office, the Air Force Research Office, and the National Science Foundation.

Source link

Previous articleInducing hibernation-like state in mice may protect organs during heart surgery – ScienceDaily
Next articleGovernor Newsom announces that the peak fire season is over in California