Chinese scientists are asserting a major AI breakthrough: the development of the world’s first “brain-like” large language model that’s far more efficient than today’s models. The new model, called SpikingBrain1.0, is designed to mimic how human neurons fire sparingly, using only the necessary neural units – and notably it can run without any Nvidia GPUs . Instead of analyzing entire sentences with heavy “attention” mechanisms, SpikingBrain focuses on a rolling window of recent words (much like a human concentrating on the latest context). This brain-inspired approach, described in an early research report, purportedly makes the AI 25 to 100 times faster than traditional models like ChatGPT while balancing efficiency and accuracy. The model runs on China’s homegrown MetaX chips, highlighting an effort to reduce reliance on US hardware . If these claims hold up, it could mark a significant shift toward more energy-efficient AI – though global experts will be eager to see peer review of the results.

Sources: The Independent