meta
Researchers have created a neuromorphic chip that can execute AI applications directly in memory without the need for a network connection to a cloud. Additionally, the chip uses less energy than other chips, making it more efficient. The discovery is anticipated to make it possible to use AI in a variety of edge devices where it can carry out a variety of complex activities independently of a centralised server.
NeuRRAM chips have shown to be both more effective than compute-in-memory chips and capable of producing results that are equally accurate. As a result, the chip may find use in processes like voice recognition and reconstruction as well as image recognition.
Power and computational aptitude are both necessary for AI computing. The majority of edge computing AI applications demand that data be sent from the device to the cloud, where processing takes place. The data is then transferred back to the apparatus. This is due to the fact that the majority of edge devices run on batteries and only have a small amount of computing power.
The NeuRRAM chip, created by experts at the University of California, lowers this power consumption, enhancing the intelligence, reliability, and usability of edge devices. Additionally, it improves data security as data privacy threats exist while sending data from the device to the cloud.
It is thought that relocating the data is a difficult task. Weier Wan, a Stanford University PhD alumnus who worked on the chip at UC San Diego, said, "It's the equivalent of doing an eight-hour commute for a two-hour workweek." He also contributed to the research that was published in Nature.
Resistive random-access memory, a type of non-volatile memory that permits computation inside a memory without needing a separate computer unit, was utilised by the team. Although the compute-in-memory approach is not new, the NeuRRAM chip stands out because it provides excellent efficiency and flexibility for a variety of AI applications while maintaining the same accuracy.
When the chip was put through a variety of tasks, researchers were able to see outstanding results that were on par with those of other digital processors.
