Engineers at Cornell University have developed the world's first microchip that contains a complete microwave neural network. This device is capable of processing information in real time using just 200 milliwatts of energy.
The main difference between this chip and conventional neural networks lies in its unique architecture. Instead of performing sequential digital operations, it processes data streams at frequencies in the tens of gigahertz. This makes it significantly faster than most existing chips. The device can be instantly reconfigured for various tasks, from deciphering radio signals to tracking radar targets.
The chip has already demonstrated high accuracy in tests, reaching up to 88% in recognizing wireless signals. Due to its sensitivity, it can be used to detect anomalies in wireless systems, making it useful for hardware cybersecurity. In the future, if its power consumption can be reduced, such a chip could be integrated into smartwatches and smartphones, allowing them to train AI models directly on the device without needing access to cloud servers.
ORIENT
