Artificial intelligence is one of the most important features of recent times. Starting from our daily uses of smartphones and other electronic devices, artificial intelligence is now everywhere to recognise and classify data for better usage and experience. The artificial intelligence that is used in calculating the algorithms is known as deep neural networks. These networks when connected to the internet, work faster and smarter to help every electronic user in finding references of their preferred area, as the deep neural networks consistently observe the user’s device functions. However, deep neural networks require strong computing power and large memory to run quickly in an average smartphone which led many researchers to find ways to accelerate the process of recognition and classification of data.
Researchers at Northeastern University presented their study on the deep neural networks that can work on smartphones or similar devices and even 56 times faster with similar accuracy at a conference on artificial intelligence this February in New York. Yanzhi Wang, an assistant professor of Electrical and Computer Engineering at Northeastern, says that it is difficult to achieve real-time execution of these neural networks on a mobile device, “but we can make most deep learning applications work in real-time.” Any deep neural network needs powerful internet access to work in a phone, which can be a problem in many situations with poor internet connection or network, so Wang and his team have come up with a way to reduce the size of the neural network model and an automatic code generating system which can work efficiently in poor network coverage areas.
Yanzhi Wang has considered the issues that artificial intelligence can probably face for it is now significantly used in medical purposes and delay of fractions of a second can make huge differences in this matter. Furthermore, a device powered by artificial intelligence can reduce privacy concerns as all the local data is shared over the neural network without sending it off to distant servers like a cloud. Wang specifically proved that these networks do not always need dedicated storage and power.
Rubena Bose