Human Learning Can Be Duplicated in Solid Matter

Researchers have found that learning — a universal feature of intelligence in living beings — can be mimicked in synthetic matter, a discovery that in turn could inspire new algorithms for artificial intelligence (AI).

-Rutgers University

Researchers from Rutgers University and its colleagues have shown that learning, a universal characteristic of intelligence in live organisms, can be replicated in synthetic matter, a finding that may inspire new algorithms for artificial intelligence (AI). Humans have the capacity to constantly learn from and adapt to changing surroundings as one of their most basic traits. However, artificial intelligence (AI) has only lately begun to stray from its limited emphasis on simulating human reasoning. Human cognition is already being mimicked in machines that can learn, remember and make choices as a human brain does, according to the latest study.

Inspiring new AI and neuromorphic computer algorithms with the flexibility to deal with uncertainties, contradictions, and other elements of daily life may be achieved by simulating such characteristics in the solid state. Neuromorphic computing uses artificial nerve systems to transmit electrical impulses that imitate brain signals to replicate the neural structure and function of the human brain. An insulating substance known as nickel oxide was examined by scientists at Rutgers University, Purdue University, and other institutions to see how its electrical conductivity altered over time.

According to Rutgers-New Brunswick postdoctoral associate Subhasish Mandal, “the aim was to develop a material whose electrical conductivity could be adjusted by changing the concentration of atomic defects using external stimuli such as oxygen, ozone, and light.” This material’s behavior when doped with oxygen or hydrogen was examined, as was how the external stimulation changed the material’s electrical characteristics. When the gas stimulus changed quickly, the material couldn’t react fully, according to the researchers. They were both unstable, and their reaction dwindled with time. When ozone, an acrid gas, was used as a stimulus, the material’s response spiked before tumbling back down.

We found common learning features such as habituation and sensitization in all living species, which is very intriguing,” Mandal added. “In turn, these material properties may serve as inspiration for future AI systems. We think that collective behavior of electrons in a quantum solid will one day inspire artificial intelligence in the same way that collective motion of birds or fish has. There is an increasing need for hardware that can host adaptive memory characteristics beyond what is utilized in current computers, he said. For future brain-inspired computer and robotics testing we found that nickel oxide insulators, which were previously confined to academic endeavors, may be intriguing possibilities.

Journal Reference: Zhen Zhang, Sandip Mondal, Subhasish Mandal, Jason M. Allred, Neda Alsadat Aghamiri, Alireza Fali, Zhan Zhang, Hua Zhou, Hui Cao, Fanny Rodolakis, Jessica L. McChesney, Qi Wang, Yifei Sun, Yohannes Abate, Kaushik Roy, Karin M. Rabe, Shriram Ramanathan. Neuromorphic learning with Mott insulator NiOProceedings of the National Academy of Sciences, 2021; 118 (39): e2017239118 DOI: 10.1073/pnas.2017239118

Categories: Tech&Innovation