They are specialized junction points that allow neurons to transmit messages to other neurons or non-neuronal cells such as muscles or glands.
Teams around the world are working on technologies designed to mimic the brain’s connectivity systems in various ways to perform tasks such as computer intuition and natural language processing. It develops more advanced artificial intelligence of the type called neural networks. At these times, a MIT teamis working on an alternative system that uses analog devices that can more effectively mimic the brain’s processes. (The findings are described in an article published in the journal Nature Communication.)
Ion-based synapse technology can provide energy-efficient simulations of the learning process of the brain for neural networked artificial intelligence systems.
Neural networks; It simulates the way learning takes place in the brain, based on the gradual attenuation or strengthening of the connections between neurons, known as synapses. The basic element of this physical neural network is the resistive switch that can electrically control electrical conductivity. This control, or modulation, mimics the strengthening and weakening of synapses in the brain. So far most of the candidate analog resistive devices have encountered some drawbacks for such simulated synapses. Either it was inefficient in terms of energy use, or it ran from one device or loop to another sporadically.
Working Logic of Sinaps New Technology
The researchers say the new system tackles both challenges. The resistive switch in this study is an electrochemical device made of tungsten trioxide that works in a manner similar to charging and discharging batteries. In this case, protons can move in and out of the crystal lattice of the material, depending on the polarity and strength of the applied voltage. These changes remain constant until they are altered by a reverse applied voltage, just as synapses are strengthened and weakened.
MIT professor Bilge Yıldız: “This process is very similar to the working principles of brain synapses. Not with protons there; calcium, potassium, magnesium, etc. we work with other ions. By moving these ions, you change the resistance of the synapses, this is an element of learning.
The materials used in the introduction of the new device were chosen for compatibility with existing semiconductor manufacturing systems. However, it contains a polymer material that limits the device’s tolerance to heat. That’s why the team is still looking for better ways to encapsulate the hydrogen source for long-term transactions.
MIT professor Bilge Yıldız says there is a lot of research to be done for this device, especially in terms of materials, and all this research will take time.