The new processor will enable you to run neural networks on smartphones
Neural networks are becoming more common day by day. But their work requires a lot of energy. And if a couple of years ago someone said that neural networks can work in ordinary smartphones, many would have thought that it was a joke. However, thanks to the efforts of experts from the Massachusetts Institute of Technology (MIT), an AI-microprocessor for portable gadgets appeared.
According to the Engadget edition, MIT specialists have developed a processor whose power consumption is 95% less than existing analogs with extremely high performance. Thanks to the new processor it will be possible to use the resource and energy-consuming applications like neural networks, even on handheld devices.
In addition to reduced power consumption, the microprocessor from MIT is more quickly used in smartphone building today 3-7 times. This was achieved thanks to the new architecture using the scalar product method, which allowed to calculate the connections at once for the whole data set without sending the intermediate results to the RAM and back. About the new chip even expressed the president of IBM Dario Gil:
“The new chip uses an extremely energy-efficient approach. This is a new word in the implementation of convolution operations with huge arrays of memory. This will make it possible to create more complex and accurate neural networks for wearable Internet devices of things. “
Neural networks (artificial neural network) – a system of connected and interacting with each other simple processors (artificial neurons). Such processors are usually quite simple (especially in comparison with processors used in personal computers). Each processor of such a network only deals with the signals it periodically receives, and the signals it periodically sends to other processors. And, nevertheless, being connected to a large enough network with controlled interaction, these processors together are able to perform quite complex tasks, because neural networks are trained in the process of work.