Experimental 'microwave brain' chip processes AI using under 200 milliwatts of power

Cornell engineers created a microwave brain chip using under 200 milliwatts, outperforming traditional processors.

: Cornell University researchers have engineered a groundbreaking chip that marries digital processing with a microwave neural network, capable of executing complex tasks faster than traditional processors while using less than 200 milliwatts. The innovative chip, named the 'microwave brain', processes data using wireless analogue communication by manipulating microwaves at gigahertz frequencies. This unique configuration enables the chip to handle AI workloads with high accuracy while vastly reducing power consumption and size, making it ideal for edge computing applications. Furthermore, funded in part by DARPA, the project exemplifies how modern AI technology continues to challenge conventional computing boundaries.

The revolutionary chip developed by researchers at Cornell University is a marriage of traditional digital processing paradigms and a microwave-based neural network architecture. Utilizing less than 200 milliwatts of power, this 'microwave brain' chip performs tasks that resemble those handled by conventional neural networks. It challenges traditional circuit designs by leveraging microwave technology and bypasses the linear instruction execution of digital hardware, thus providing significant improvements in speed and power efficiency.

The microwave brain chip processes data via analogue wireless communication and engages in real-time frequency-domain computation. This makes it ideal for tasks like decoding radio signals, tracking radar targets, and processing digital data. As it interacts directly with incoming data inputs, it can also detect anomalies in wireless communications across a range of microwave frequency bands. Such capability would position the chip as a promising tool for cybersecurity and defense applications, given its efficiency and adaptability.

By integrating waveguides into the neural network through a probabilistic design strategy, the researchers succeeded in managing AI workloads effectively. The design allows the microwave brain to classify wireless signal types with a precision of at least 88 percent, using far less energy than its digital counterparts, all within a much more compact hardware footprint. This efficiency could lead to the chip's incorporation into edge computing devices, namely smartphones and wearables, reducing their cloud dependency and expanding their AI capabilities.

Image of Microwave Brain Chip

The achievements and potential applications of this chip were also fueled by funding from DARPA, Cornell, and the National Science Foundation. Published in the August 14 issue of Nature Electronics, this project highlights the futuristic potential of AI-enabled devices such as smartphones, smartwatches, and glasses. While tech giants like Apple and Meta pursue AI advancements, the exploration of neural networks for wearables remains underdeveloped, presenting unique opportunities for innovation.

The microwave brain chip is still experimental, its scalability and accuracy improvements in development as researchers seek ways to expand its usage across diverse platforms. With minimal power consumption and size reduction advantages, future devices equipped with such technologies could redefine the landscape of portable AI-driven technology.

Sources: Cornell University, TechSpot

AIinnovationMachine Learning