Prototype technology shrinks AI to deliver brain-like functionality in a powerful device.
Researchers have developed artificial intelligence technology that combines images, processing, machine learning and memory into a single light-powered electronic chip.
The prototype shrinks artificial intelligence technology, mimicking the way the human brain processes visual information. The nanoscale advance combines the basic software needed to power artificial intelligence with image capture hardware in an electronic device.
With further development, the light-controlled prototype can enable more intelligent and smaller autonomous technologies such as drones and robotics, plus intelligent wearables and bionic implants such as artificial retinas.
The study, by an international team of Australian, American and Chinese researchers led by RMIT University, was published in the journal Extended materials.
Lead researcher Associate Professor Sumeet Valia of RMIT said the prototype provides brain-like functionality in a powerful device.
“Our new technology is radically increasing its efficiency accuracy by combining multiple components and functionalities into one platform, ”said Valia.
“This brings us closer to a whole device with artificial intelligence, inspired by the greatest computer innovation in nature – the human brain.
“Our goal is to reproduce a basic characteristic of how the brain learns by printing vision as memory. The prototype we developed is a major leap forward in neurorobotics, better human-machine interaction technologies, and scalable bionic systems. “
Common package: advanced AI
Artificial intelligence usually relies heavily on software and off-site data processing. The new prototype aims to integrate electronic hardware and intelligence together for quick on-site solutions.
“Imagine a dash camera in a car that integrates with our neuro-inspired hardware – meaning it can recognize lights, signs, objects and make instant decisions without having to connect to the Internet,” Wahlia said. , who manages Functional Materials and Microsystems Research Group at RMIT, said.
“By combining everything into one chip, we can provide unprecedented levels of efficiency and speed in autonomous and AI-driven decision-making.”
The technology is based on an earlier prototype chip from the RMIT team, which uses light to create and modify memories.
The new built-in features mean that the chip can now capture and automatically enhance images, classify numbers and be trained to recognize patterns and images with over 90% accuracy.
The device is also easily compatible with existing electronics and silicon technologies, for seamless integration in the future.
Seeing the light: how technology works
The prototype is inspired by optogenetics, a new tool in biotechnology that allows scientists to delve into the body’s electrical system with great precision and use light to manipulate neurons.
The AI chip is based on an ultra-thin material – black phosphorus – which changes the electrical resistance in response to different wavelengths of light. Various functionalities such as images or memory storage are achieved by illuminating different colors of light on the chip.
Lead author of the study, Dr. Taimur Ahmed of RMIT, said that light-based calculations were faster, more accurate and required much less energy than existing technologies.
“By packing so much basic functionality into one compact nanoscale device, we can expand the horizons of machine learning and AI to be integrated into smaller applications,” Ahmed said.
“Using our artificial retina chip, for example, would allow scientists to miniaturize this emerging technology and improve the accuracy of the bionic eye.
“Our prototype is a significant advancement in the best of electronics: a brain on a chip that can learn from its surroundings just like us.”
This work was carried out in part at the Micro Nano Research Facility (MNRF) at RMIT, with the support of the RMIT Microspy and Microanalysis Research Facility (RMMF), the National Computational Infrastructure Australia (NCI), the Multimodal Australian Sciences Imaging and Visualization Environment (MASSIVE) and Pawsey Supercomputing Facility.
Reference: “Fully controlled by light memory and neuromorphic calculations in layered black phosphorus” by Taimur Ahmed, Mohamed Tahir, May Xian Lowe, Yanun Ren, Sheriff Abdulkader Tawfik, Edwin L.H. , Madhu Bhaskaran, Sharath Sriram and Sumeet Walia, 17 November 2020, Extended materials.
DOI: 10.1002 / adma.202004207