Chinese team releases brain-like AI model, with long-task operation speed increased 100 times

Foreign Media: Chinese Team Launches Brain-like AI Model, Running Long Tasks 100 Times Faster

A team from the Institute of Automation, Chinese Academy of Sciences, has released the world's first "brain-like" large language model, SpikingBrain 1.0, aiming to reduce energy consumption and reduce reliance on NVIDIA chips. The system simulates the mechanism of the human brain, activating only necessary neurons, unlike mainstream models like ChatGPT that require the entire network to be activated, thus saving energy and accelerating response time.

Research shows that the model can learn with less than 2% of the training data required by traditional systems, and it remains efficient when processing extremely long texts. In some tests, its operation speed is up to 100 times faster than traditional models.

SpikingBrain 1.0 runs entirely within China's independent AI ecosystem, using the MetaX chip platform instead of NVIDIA GPUs. In the context of tightened U.S. export controls on high-end AI chips, it holds significant strategic importance.

Original article: www.toutiao.com/article/1842888901445699/

Statement: This article represents the views of the author.