A research team from Ecole Polytechnique Federale de Lausanne (EPFL) has announced a breakthrough in developing a miniaturized artificial intelligence chip that enables direct decoding of brain signals. Dubbed as miniaturized BMI (MiBMI), this chip boasts 192 input channels and houses 31 sets of neural decoders. It has the ability to accurately recognize hand gestures with a high precision of 91.3%, although it has not yet been tested for actual brain connectivity in humans. However, the team has successfully decoded sounds heard by mice, classifying 6 types of sounds with an accuracy of 87%.
Utilizing 65-nanometer manufacturing technology, the chip occupies a compact area of 2.46 square millimeters and consumes 883 microwatts of power. When combined, the chip area for data storage and processing amounts to 8 square millimeters.
Source: EPFL
TLDR: EPFL research team introduces MiBMI, a miniaturized AI chip for decoding brain signals with high accuracy. Tested successfully on mice, the chip uses advanced technology for compact design and efficient power consumption.
Leave a Comment