Neural Networks To Devices Introduced By Mediatek

In the past two years, the industry has made immense advancement in artificial intelligence using artificial neural networks, known as deep learning. With huge amounts of data, software framework, and processing resources, a network of filters is created and optimized to execute selected functions such as image recognition. As the neural network gets trained, it creates models that can later be utilized by computer solutions with far fewer processing resources to perform the preferred function with equivalent data. This is commonly referred to as an inferior solution, which is common processing element similar to CPUs and GPUs or other custom processing solutions.

Most artificial intelligence applications, such as Amazon Echo with Alexa, handle all or part of the processing in the cloud. While this is pleasing for solutions that are always connected to power and a network, there are many other applications, especially consumer applications that need local processing. Local processing also provides supplementary data security as it is not sent over a network and is accessible to others. Like many others in the industry, MediaTek seeks to bring local neural computing in reality.

Combined with MediaTek’s CorePilot technology, the various computer resources are efficiently utilized to carry out the obligatory tasks. Preferably, the local solution should be able to learn and acclimatize the models that are used, but it seems that the MediaTek platform, such as most other device platforms, permits it to learn on the device without any assistance on cloud-based resources. With so many sources on a single chip, learning on the device is not far away.

The MediaTek AI platform, which consists of the Helio X family of mobile SoCs and the Deep Learning SDK, is a significant step toward device-level intelligence and is becoming a consumer-centric, giving New AI levels right in their hands.