INFORMATION TECHNOLOGY

Scientists build a brain-like cognitive intelligence engine “Zhimai” of full-pulse neural networks


The brain-inspired cognitive intelligence engine (abbreviated as BrainCog, Chinese known as “Zhimai”) led by researcher Zeng Yi of the Institute of Automation of the Chinese Academy of Sciences has built a full-pulse neural network brain-inspired cognitive intelligence engine (abbreviated as BrainCog, name “Zhimai”), and carried out comprehensive open source and open-source to provide basic support for exploring brain-inspired intelligent research for general artificial intelligence. Facilitate the exploration of the computational nature of natural intelligence and the development of a new generation of artificial intelligence. The article was published as a cover article in the journal Patterns of Cell Press.

Based on the principle of multi-scale bioplasticity, “Zhimai” supports full-pulse neural network modeling, brain-inspired artificial intelligence models and brain function and structure simulation capabilities, and provides a relatively complete and systematic set of interface components for researchers of brain-like artificial intelligence and computational neuroscience, including neuronal computing models with different degrees of fineness, rich brain-like learning and plasticity laws, different neural network connection modes, diverse coding methods, rich functional brain region models and software and hardware collaborative systems.

At present, “Zhimai” has made substantial progress in perception and learning, knowledge representation and reasoning, decision-making, embodied intelligence, social cognition, and developmental evolution. In terms of software and hardware collaboration, the team developed and released the FPGA-based brain-like pulse neural network accelerator “BrainCog Firefly”, which improves the inference performance of the pulse neural network on edge devices from the hardware level, and has great application potential in real tasks such as intelligent vehicle autonomous visual positioning and navigation based on pulse neural network, high-speed obstacle avoidance of UAV based on event camera, robot environment exploration and complex task self-adaptation collaboration. Based on the “Zhimai” framework, the team also developed the pulsed neural network artificial intelligence engine “BORN” dedicated to obtaining general cognitive abilities, integrating perception and learning, working memory, long-term memory, knowledge representation and reasoning, decision-making, motor control, attention and awareness, empathy, social cognition and other brain cognitive functions.

“Zhimai” has preliminarily integrated the team’s early self-based mirror test, rubber hand experiment, psychological speculation, and cognitive simulation of emotional empathy, and further improved the self-organized learning ability of brain-like agents with the artificial intelligence engine BORN as the core. In terms of the collaboration of different learning models, BORN takes “Zhimai” as the basic computing platform to initially realize the composition and performance of music based on humanoid robots and emotions driven by humanoid robots that integrate multiple cognitive functions such as perception and learning, sequence learning and generation, knowledge representation and reasoning, and motion control, and support humanoid robots to achieve visual emotion recognition, emotion-dependent music creation and music performance.

In the future, the team will continue to explore and improve the general artificial intelligence engine based on pulsed neural networks to provide strong foundation support for research in the fields of brain-like artificial intelligence and neuroscience. (Source: Institute of Automation, Chinese Academy of Sciences)

Related paper information:https://doi.org/10.1016/j.patter.2023.100789

Computing components and applications of BrainCog, a brain-like cognitive intelligence engine “Zhimai”

Patterns journal cover

Special statement: This article is reproduced only for the need to disseminate information, and does not mean to represent the views of this website or confirm the authenticity of its content; If other media, websites or individuals reprint and use from this website, they must retain the “source” indicated on this website and bear their own legal responsibilities such as copyright; If the author does not wish to be reprinted or contact the reprint fee, please contact us.



Source link

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button