Mr. Binder, where are the computers used?

Raphael Binder: They perform decentralized AI tasks and are therefore used as edge computers. This means that they are not typically located in an air-conditioned server room, but are used in the field - be it in vehicles, machines or production halls. Accordingly, they meet high requirements in terms of robustness and reliability.

What would be typical areas of application for the computer in production?

Raphael Binder: We are currently tackling the first deep learning applications with various customers from the mobile computing sector. This is mainly because Syslogic is very strongly positioned in the field of vehicle computers. However, we also see many applications for AI embedded computers in manufacturing - for example in robotics, where the aim is to coordinate motion sequences using AI, or in automated guided vehicles, which can also be used in unsupervised areas in the future thanks to AI.

Why is the new computer particularly suitable for AI applications?

Raphael Binder: The basis of our AI embedded computer is the Jetson TX2i module from Nvidia. This is characterized by an efficient and powerful quad-core processor platform. At the heart of the Jetson-TX2 is the ARM SoC Tegra X2 called Parker. It combines two computing cores with the Denver 2 microarchitecture developed by Nvidia itself with four Cortex-A57 cores and a Pascal GPU. The latter has 256 shader cores. This means that our AI computers are ideally equipped for AI applications.
In addition to the actual processor platform, Nvidia also boasts the JetPack Software Development Kit, which contains CUDA libraries, programming interfaces and examples. This makes it much easier for users to get started with AI applications.

What are CUDA applications about?

Raphael Binder: Compute Unified Device Architecture - CUDA for short - is a programming technology developed by Nvidia that allows a computer's GPU to be used for parallel calculations. It has already become widely established and a large number of libraries and developer tools are available, making it easier for users to implement AI applications.

Why AI directly on computers and not in the cloud?

Raphael Binder: Data must be evaluated on site and decisions made immediately. For applications in autonomous or semi-autonomous vehicles, it is not possible to wait for results from the data center. In addition, an edge computer does not always have a connection to the cloud, especially in
connection to the cloud. However, it is obvious that the artificial intelligence of entire data centers cannot be managed with low-power embedded systems. Accordingly, our embedded systems used as edge computers perform limited functions. The resource-intensive calculations take place downstream in the cloud.