New Silicon Transistor Able to Mimic Neurons and Synapses | Generative ai google search | Google ai course for beginners | Generative ai course google | Turtles AI

A research team from the National University of Singapore has developed a silicon transistor that can mimic both neurons and synapses, improving the efficiency of AI systems.

Key points:

• A single silicon transistor can mimic the behavior of neurons and synapses.
• This allows for a significant reduction in the size and cost of neuromorphic circuits.
• The discovery exploits the phenomenon of impact ionization, traditionally seen as a defect.
• The device was built using 180-nanometer technology, which is accessible and well-established.

The team led by Associate Professor Mario Lanza, of the Department of Materials Science and Engineering at the College of Design and Engineering at the National University of Singapore, has developed a silicon transistor that can mimic both the behavior of neurons and synapses. This innovation could improve the efficiency of AI systems by reducing the size and cost of neuromorphic circuits. Traditionally, building electronic neurons and synapses requires the interconnection of numerous transistors, increasing the complexity and power consumption of the devices. The solution proposed by Lanza’s team exploits the phenomenon of impact ionization, generally considered a failure mechanism in silicon transistors. By controlling this phenomenon through the regulation of the resistance at the ground terminal, the transistor can generate current peaks similar to neuronal activation or maintain persistent resistance states, simulating synaptic behavior. A relevant aspect of this research is the use of transistors with 180 nanometer technology, a consolidated and accessible platform, which facilitates industrial integration without the need for advanced manufacturing processes. Professor Lanza emphasizes that, once the operating mechanism is understood, the implementation is mainly based on microelectronic design. Dr. Sebastián Pazos, first author of the study, highlights how this approach represents an alternative to the traditional miniaturization of transistors, promoting a more efficient and economically sustainable computing paradigm.

This research opens new perspectives for the development of advanced neuromorphic systems, capable of processing information more efficiently and with lower energy consumption.