A team of scientists from China and Switzerland has collaborated to develop a chip that simulates the human brain, focusing on energy-efficient sensor computing and mimicking neurons and synapses.
The Speck chip consumes very low power at rest, only 0.42 milliwatts. (Image source: dimsumdaily.hk).
The human brain serves as a “command center” that regulates and controls all bodily functions. Together with the spinal cord and nerves, it forms the central and peripheral nervous systems. The nervous system operates by transferring electrical signals back and forth from the brain to the spinal cord, nerves, and various parts of the body. It is estimated that these activities consume around 20 watts of power, which is significantly lower than current artificial intelligence (AI) systems. Therefore, brain-simulating computing offers a promising energy-efficient solution for AI.
According to a recent study published in the journal Nature Communications, researchers from the Institute of Automation at the Chinese Academy of Sciences and SynSense AG Corporation in Switzerland have developed this asynchronous chip, named Speck, which consumes an extremely low power of only 0.42 milliwatts when idle, meaning it almost does not consume energy when there is no input information.
By simulating the dynamic imbalance characteristics of neural spike networks in the brain, the research team created an attention-based tool, where strong external stimuli typically receive greater attention from the brain.
This tool can meet the algorithmic requirements of dynamic computing, achieving a real-time power consumption of 0.70 milliwatts.
Li Guoqi, one of the study’s authors, stated that this research provides AI applications with an intelligent brain-simulating solution characterized by exceptional energy efficiency, minimal latency, and reduced power consumption.
