A group of developers at Microsoft embodies the idea of ??changing the paradigm of artificial intelligence expressed by the company's CEO at the recent conference Build 2017. The essence of the change lies in the transition from the cloud-centric concept to the combination of the cloud and the clever "front edge".

Devising devices that form the boundary of the electronic world, artificial intelligence, you can eliminate their dependence on the availability of a network connection, reduce traffic, reduce power consumption.

As an example, the source cites a solution based on Raspberry Pi 3, trained to recognize the protein and discourage them from flower buds and bird feeders. The scope of microsystems with artificial intelligence is very wide - from sensors of the Internet of things and portable electronics to medical devices.

The peculiarity of these devices is the presence of very modest computing resources and a small amount of memory.

To adapt the algorithms of artificial intelligence to the specified hardware capabilities, researchers apply two approaches: they develop compression algorithms for machine learning models that allow them to be transferred to microsystems, and create sets of highly specialized algorithms that are optimal only for a specific task. As claimed, this allows you to reduce the requirements hundreds and thousands of times.

The main difficulty in this case is to keep the accuracy of the algorithms at a level sufficient for practical application.

Intermediate results of the team of Microsoft experts are available on the GitHub website. The developers hope that in the future, artificial intelligence tools will be widely available to all enthusiasts who use Raspberry Pi and other similar systems in their projects.