IIT teams create ultra-thin, light-sensitive chip that cuts visual AI energy use by 500-fold
Researchers from IIT Guwahati and IIT Kharagpur have built an ultrathin, light-responsive chip that merges sensing, memory and processing while using 500 times less energy than comparable devices.
A collaboration between IIT Guwahati and IIT Kharagpur has produced an ultrathin electronic component that detects light, retains information and processes it on a single platform, consuming 500 times less energy than standard MXene-based devices. The researchers demonstrated the chip’s capability with facial-recognition tests achieving 90% accuracy and traffic-scene segmentation at 85% accuracy. Built from MXene nanosheets coupled with the organic semiconductor P3HT, the interface creates a lasting electrical charge after light exposure, and a second light pulse boosts the response by 180%, emulating repeated-stimulus learning.
Lead author P K Giri explains that this integration mirrors how the human brain combines perception with memory, enabling rapid recognition without exhaustive pixel-by-pixel analysis. Co-author Chandra Sekhar Tiwary notes the potential for substantial energy savings in applications such as autonomous navigation, surveillance and industrial inspection. The findings, published in the journal Carbon, add to ongoing efforts in neuromorphic computing to overcome the von Neumann bottleneck in AI hardware.
Why it matters
The chip could dramatically lower power consumption for AI vision tasks, enabling more efficient and sustainable smart devices.
In this story
