The human brain simulation chip consumes almost no energy when there is no input information

A group of Chinese and Swiss scientists collaborated to develop a human brain simulation chip called Speck. This chip simulates human brain neurons and synapses, consuming extremely low power at rest, only 0.42 milliwatts.

The human brain simulation chip consumes almost no energy when there is no input information Picture 1 The human brain simulation chip consumes almost no energy when there is no input information Picture 1

The human brain is an "executive center" that governs and controls all functions of the body. The brain, spinal cord, and nerves combine to form the central and peripheral nervous systems.

The process of relaying electrical signals back and forth from the brain to the spinal cord, nerves and to different parts of the body helps the nervous system operate.

It is estimated that these activities consume about 20 watts of electricity in total, a number that is too small compared to the amount of electricity that current artificial intelligence (AI) systems consume. Therefore, scientists say that brain-simulating computing offers promising energy-efficient AI.

The team of scientists simulated the dynamic imbalance properties of the brain's impulse neural networks to create an attention-based tool in which strong external stimuli often receive more attention. Pay more attention to the brain.

The engine is capable of meeting the algorithmic requirements of dynamic computing, and consumes only 0.70 milliwatts of real-time power.

Li Guoqi, one of the authors of the study, said the human brain simulation chip could be used for AI applications, providing a smart solution that saves energy, minimizes latency and reduces consumption. absorb electricity.