Researchers "think" robots can be like humans

SAN FRANCISCO - Jan. 7, 2022 - PRLog -- These nerve cells were created from living cells and served as a physical reservoir for the computer to use in order to build coherent signals. These findings imply that by giving disruption signals to an embodied system, goal-directed behaviour can be generated without any extra learning.

Researchers from the University of Tokyo describe how a robot could be trained to walk around a maze by electrically stimulating a culture of brain nerve cells attached to the machine in Applied Physics Letters, published by AIP Publishing. These nerve cells, or neurons, were produced from living cells and served as the computer's physical reserve for constructing coherent messages. The signals are considered homeostatic signals, as they inform the robot that its internal environment is kept within a particular range and serve as a baseline as it moves freely through the maze.

An electric impulse was sent to the neurons in the cell culture if the robot went in the wrong direction or looked the wrong way. Throughout the trials, the robot was fed homeostatic impulses that were interrupted by disruption signals until it performed the maze problem correctly. These findings imply that by giving disruption signals to an embodied system, goal-directed behaviour can be generated without any extra learning. Because the robot couldn't see its surroundings or get any other sensory information, it was completely reliant on electrical trial-and-error impulses.

"Our findings encouraged me to speculate that intelligence in a living system develops through a method deriving a coherent output from a disorderly state, or a chaotic state," said co-author and associate professor of mechano-informatics Hirokazu Takahashi. Using this principle, the researchers demonstrate how physical reservoir computers can extract chaotic neural impulses and transmit homeostatic or disturbance signals, resulting in intelligent task-solving abilities. As a result, the computer generates a reservoir of knowledge about how to complete the task.

"A brain of an elementary school student is unable to solve mathematical problems in a college entrace exam, possibly because the mechanics of the brain or their 'physical reservoir computer' is not rich enough," Takahashi explained. "How rich a repertoire of spatiotemporal patterns the network can generate determines task-solving abilities."

Physical reservoir computing, the team hopes, will aid in a better understanding of the brain's operations and could lead to the construction of a revolutionary neuromorphic computer.

More of the latest news in investment, innovation and technology you may also visit us at HyperCharge Technology through our website https://hypercharge-tech.com/ or contact us via email.