A Chip That Can “See” and Store Memories

by | 16. Jun 2023 - 13:29 | Technologies

The use of indium oxide enables real-time decision-making, researchers say, and could advance application fields such as autonomous driving.

Australian researchers have developed a chip that can mimic human vision and memory by capturing, processing, and storing optical information, according to a statement from the Royal Melbourne Institute of Technology (RMIT), which is leading the project.

The underlying technology belongs to the field of neuromorphic engineering, a method of replicating the way natural nerve cells (neurons) work. Neuromorphic vision systems are based on analog information processing similar to that in the human brain. Compared to current digital technologies, visual tasks could be performed faster, more efficiently, and with significantly lower energy consumption this way, explains first author and RMIT doctoral student Aishani Mazumder.

With the newly developed chip, the research team says it has now mastered a major challenge: It succeeded in accommodating the sensor, data processing, and memory functions in a single small device. Time- and energy-intensive devices such as data transmission to external processors and the processing of irrelevant information are no longer required – decisions can be made in real-time, says team leader Professor Sumeet Walia. This is made possible by the technology metal indium, which is otherwise used primarily for touchscreens, LEDs, and fiber optic cables and is integrated here in the form of a wafer-thin layer of doped indium oxide. This precise design enables the chip to mimic the human eye’s ability to capture light, package and transmit information like an optical nerve, and store and classify it in a memory system similar to the brain, according to the researchers.

Chip Could Be Used in Self-Driving Cars and Autonomous Surgery

The invention could become significant for application fields that require fast and complex decision-making capabilities, such as self-driving cars that can see and recognize objects on the road in the same way as human drivers, the scientists said. Bionic vision, autonomous operations in dangerous environments, and shelf-life assessments of food are also conceivable applications. Unlike conventional vision systems, which are programmed according to specific rules, neuromorphic systems can also adapt to new situations over time and become more efficient as they gain experience, Walia said.

Photo: iStock/CreativaImages

AD