Using AI efficiently and in a sustainable way

Artificial intelligence is increasingly being run directly on devices, machines, and sensors to reduce latency, meet data privacy requirements, and decrease reliance on cloud infrastructures. At the same time, limited computing resources and energy availability place high demands on the efficiency of AI systems.

The workshop provides an overview of methods and technologies for energy-efficient AI at the edge. Building on classical approaches such as model pruning, quantization, binary neural networks, and resource-efficient network architectures for embedded systems, the workshop presents current strategies for reducing memory requirements, computational effort, and energy consumption. In addition, the limitations of today’s deep learning methods will be examined, and their implications for future AI systems will be discussed.

Building on this, the workshop introduces the fundamentals of neuromorphic computing. Participants will learn about the principles of event-driven information processing, spiking neural networks, and neuromorphic hardware, and will gain insight into current application areas as well as the technological maturity of this promising technology.