Created a neuromorphic chip that triggers the AI right in the memory of the device

Created a neuromorphic chip that triggers the AI right in the memory of the device

Newswise International has announced the creation of a neuromorphic neurram chip that calculates in memory. Its energy efficiency is twice as high as that of existing chips, so it can be used for the autonomous operation of IE applications.

Most applications in smartphones or other peripheral devices transmit data to the cloud where they are analysed and processed by the AI. The result is returned to the device because most peripherals feed from the battery and, as a result, have a limited capacity that can be explained for calculation by the developers.

To solve the problem of data transmission, scientists used resistive memory with arbitrary access . It is an energy independent storage device in which calculations can be made directly in memory rather than in separate computing units.

In order to overcome these limitations, scientists have developed a methodology that takes into account various limitations ranging from the physics of memory devices to diagrams and network architecture.

A study of the device has shown that it has twice as high energy efficiency and 7 to 13 times as high computing density as modern chips. In addition, 99% accuracy is achieved when using IE chip applications when recognizing numbers, 85.7% when classifying images, and 84.7% when using Google voice commands.