Computers have made tremendous advancements in terms of their power and capabilities, often surpassing human brains in tasks such as data storage, analysis, and communication. However, one area where human brains still outperform computers is in energy efficiency. According to UC Santa Barbara electrical and computer engineering Professor Kaustav Banerjee, the most efficient computers are still significantly more energy-intensive compared to the human brain, particularly in tasks like image processing and recognition. This energy inefficiency is a pressing issue, especially considering the increasing global energy consumption by on-chip electronics, fueled by applications like artificial intelligence.
To address this energy efficiency gap, neuromorphic (NM) computing has emerged as a promising solution. By mimicking the structure and operations of the human brain, where processing occurs in parallel across low power-consuming neurons, NM computing aims to achieve brain-like energy efficiency. In a recent paper published in the journal Nature Communications, Banerjee and his team proposed an ultra-energy efficient platform using 2D transition metal dichalcogenide (TMD)-based tunnel-field-effect transistors (TFETs). This platform could potentially reduce energy requirements to within two orders of magnitude with respect to the human brain.
The concept of neuromorphic computing has been around for decades, but recent advances in circuitry have enabled researchers to explore new possibilities for brain-inspired computing. The team’s 2D tunnel-transistors, developed as part of Banerjee’s research efforts to create high-performance, low-power consumption transistors, offer promising characteristics for neuromorphic computing. These atomically thin, nanoscale transistors are responsive at low voltages and can mimic the energy-efficient operations of the human brain. Additionally, the 2D TFETs exhibit lower off-state currents and a low subthreshold swing (SS), indicating efficient switching and lower operating voltages.
Neuromorphic computing architectures are designed to operate with sparse firing circuits, mimicking the brain’s ability to fire only when necessary. This contrasts with the traditional von Neumann architecture of current computers, where data processing is sequential, and memory and processing components are separate, leading to continuous power consumption. Event-driven systems like NM computers only activate when there is input to process, distributing memory and processing across an array of transistors. Companies like Intel and IBM have already developed brain-inspired platforms with billions of interconnected transistors, resulting in significant energy savings.
In conclusion, the development of energy-efficient computing technologies, such as neuromorphic computing, is crucial in addressing the growing energy consumption of on-chip electronics and mitigating the impact of global warming. By leveraging innovative transistor technologies like 2D TFETs, researchers are paving the way for more efficient and sustainable computing solutions that can rival the energy efficiency of the human brain.