A team of scientists from the Jawaharlal Nehru Center for Advanced Scientific Research (JNCASR) in Bangalore has developed a device that can simulate the cognitive behavior of the human brain and is more effective than traditional techniques. The new artificial intelligence technology has improved the calculation speed and power consumption efficiency. The team used a simple self-training method to create an artificial synaptic network (ASN) similar to a biological neural network.
Scientists say that the structure of the device will form on its own when heated. In order to develop synaptic devices for neuromorphic applications, the team also explored a material system that simulates the connection of neuronal bodies and axon networks similar to biological systems. Scientists say that the human brain is composed of 100 billion neurons, including axons and dendrites.
These neurons are connected to each other through axons and dendrites, forming huge connections called synapses. Scientists believe that this complex biological neural network produces excellent cognitive abilities, and added that software-based artificial neural networks (ANN) defeated humans in games such as AlphaGo and AlphaZero, and even helped manage COVID- 19 situation.
The researchers said that although the high-power von Neumann computer architecture reduces the performance of the ANN due to the available serial processing, the brain uses parallel processing to complete the work, consuming only 20W of power. Islands and nanoparticles with nanogap gaps are similar to biological neurons and neurotransmitters, where dehumidification is the process of breaking down continuous films into discrete/isolated islands or spherical particles.
“With this architecture, various high-level cognitive activities can be simulated,” the team explained in a press release from the Ministry of Science and Technology. They said that using programmed electrical signals such as ar as real-world stimuli, this hierarchical structure simulates various learning activities, such as short-term memory (STM), long-term memory (LTM), enhancement, depression, associative learning, and Interest learning, supervision, supervision, and printing, etc.
Not only that, the scientists also imitated synaptic fatigue after over-learning and self-recovery, and surprisingly, these behaviors were simulated in a unique material system without the help of external CMOS circuits. A kit that mimics the behavior of Pavlov’s dog shows the device’s potential for neuromorphic artificial intelligence.
This is a remarkable achievement, because the JNCASR team has taken another step in the realization of advanced neuromorphic artificial intelligence by organizing a nanomaterial similar to biological nerve materials.