American researchers develop computer chips to simulate human brain work

 

Using computer chips to simulate how the human brain works may reveal human cognitive capabilities.

Kwabena Boahen's laboratory at Stanford is spotless. Unlike the General neural system laboratory, there are no scattered suction tubes and piles of chemical bottles. The laboratory was replaced by a single circuit board with a special chip. The transistor is used for high-speed computation in a typical computer chip, but these seemingly cluster of micro-transistors are used to simulate the electrical properties of neurons. The transistor is designed to mimic the functional behavior of retina cells, cochlear cells, and even human haima cells (a unit located deep in the brain to classify and store information.

Although Boahen's laboratory is small in size, it has developed rapidly. Scientists and engineers use a method they call neuromorphing to build complex electronic circuits that mimic neural circuits. Their work draws on a large number of brain.

Professor Rhodes Douglas from the Zurich Institute of Neurology and information said, "With new breakthroughs in technology and theory, we should be able to explore how the brain works. These problems cannot be solved by using the most advanced digital devices. One of the research methods is to build a hardware model to simulate the work of neural circuits ."

The most fascinating thing is the brain's memory capabilities, which neuroscientists have spent decades studying. This kind of ability seems to be rooted in the hippocampus in the human brain, because if it is damaged, it will lead to amnesia.

Deep research into the hippocampus and other parts of the brain has partly revealed how neural behavior creates memories. Neurons encode the information difference in the form of electrical pulse, which can be transmitted to other neurons. When two connected neurons frequently trigger firing, the connection between them is enhanced. Therefore, the first discharged neuron can trigger the subsequent neuron discharge. As far as neuroscientists know, this process occurs on multiple adjacent neurons and produces a network that connects different neurons and encodings and information connections.

To better understand how neurons work, Boahen and graduate student John Arthur developed a chip based on the haima A3 layer. Cache is located between two other neuron layers, one of which receives information from the cerebral cortex, and the other transmits the returned information. cache is considered to be the place where memory occurs, information in this area is stored and associated.

Each circuit unit model is composed of a cluster of transistors to simulate the activity of neurons. Silicon units are arranged into 32*32 arrays, each of which is programmed to maintain a simple connection with the 21 surrounding units. At the beginning, connections between neurons were closed, simulating "silent synapses, a type of SYN that only has a SYN structure and has no information transmission function, are usually called Silent syncs. Under certain conditions, this silent SYN can be converted to a functional SYN, which may be related to the basic principles of learning and memory. Studies have found that activities that increase the number of anterior neurons can rapidly convert this silent SYN to a functional syn .)

However, Boahen explains that the chip has the ability to change the degree of neuron connection and mimic what happens When neurons work.

Arthur demonstrated the entire simulation process on the computer, and the chip remembered the previous input.

Stanford University researchers plan to increase the chip circuit to mimic the tooth-shaped seahorse layer. This is more complex, Boahen said: "We want to give it a, which can arouse memories of the entire alphabet ."

The Team also developed other "Neural form" chips that recently ambitious plans to build cerebral cortex models. The intricate structure of cerebral cortex enables us to perform complex operations. The first generation of the model will include a circuit board containing 16 chips, each containing a 256x256 silicon neuron array.

Boahen hopes that these chips can better understand how the brain works and ultimately help design neural limbs such as artificial retina.

These researchers not only need knowledge about neural systems, but also how to design chips. This research is at the forefront of the times.