Neurogrid - System that simulates the activity of the human brain

Stanford University scientists have created a system of hardware called human brain-based neurogrid with the ability to simulate the activity of 1 million neurons and billions of synaptic connectors in real time. . However, the system only has the same energy consumption as a regular tablet. The results of the manufacturing process were published in the IEEE Proceedings of the IEEE on April 28.

Currently, creating systems with the ability to simulate brain activity is not a simple matter. Personal computers can only simulate mouse cortical activity at speeds slower than 9000 times and need to consume more than 40,000 times more energy than they actually are. Therefore, with The Human Brain project , Stanford scientists have set a goal to simulate the activity of the human cortex . Initially, the energy required to operate the human brain simulation system is equivalent to the energy consumption of 250,000 households in a city.

In the study, scientists created a system called Neurogrid , consisting of 16 "Neurocore" chips that are integrated together on a circuit board. Each Neurocore has the power equivalent to 65,536 neurons. That means that the circuit board can simulate the operation of more than 1 million neurons with billions of synaptic connections operating in real time.

With such amazing capabilities, the whole system only needs 3W of power to operate. This is a very impressive breakthrough because compared to before, a computer wanted to simulate the activity of the number of neurons equivalent to thousands of watts of power, but the speed was 9,000 times slower than reality.

Currently, the biggest drawback of the human brain simulation system is the high cost of fabrication: each Neurogrid costs up to 40,000 US dollars. However, the team admits that they are using manufacturing technologies of 15 years ago. The team said that if they switched to more advanced manufacturing methods, the cost of production for each system would be reduced to about $ 400. Not only that, the application of new techniques also allows the team to increase the memory of each Neurocore to meet a higher working intensity.

The success of the research has opened up new perspectives on the technology of manufacturing robots and human brain simulations. Scientists hope that systems with similar capabilities but affordable prices can be produced to widely serve scientists' research without much knowledge about functioning. brain movement.

Although compared to brain simulation systems so far, this is truly an amazing success, but the system is still far behind the real human brain. According to Associate Professor Kwabena Beahen at Stanford, the human brain can simultaneously transport a network of nerve cells 80,000 times more, but consumes only three times as much energy as a neurogrid.

With the initial success of the project, the researchers hope that the model of human regeneration with amazing speed and performance can be applied to control artificial limbs for people with disabilities. This is the most valuable application that scientists want to achieve in the near future.