The researchers from the Institute of Photonics, Austria have created an artificial eye on a small chip by using light-sensing electronics with a neural network. This new design, presented in Vienna, Austria’s capital, has been reported to identify an object within a few nanoseconds, which is a serious advancement in the technology. 

Previous technologies would have to process a large amount of visual data, irrespective of its use for the program, thereby slowing machines down. This would prove to be problematic while executing technology like industrial robots and especially driverless cars. However, the researchers at the Institute of Photonics have reportedly created a chip that is only a few atoms thick out of Tungsten diselenide, a semiconductor with a layered crystal structure. The chip is etched with light-sensing diodes that are connected to form a neural network. Due to Tungsten diselenide’s properties, the chip’s diodes can be tweaked externally. This allows users to adjust the diode sensitivity to identify the right objects, thereby training the neural network to be trained to identify certain visual information. As an experiment, the scientists trained the chip to successfully identify various versions of letters n,v, and z. This success opens the possibilities of embedding AI into hardware that is efficient and quick. 

Since the technology is at its nascent stage, it can currently recognize block images of no more than 3x3 pixels. Currently, the chip can do a few machine-learning tasks such as classifying and encoding letters. However, the scientists are confident that scaling the model will be a straightforward process.