Sustainable AI For Energy-efficient Systems (SAFES)

Artificial intelligence (AI) and machine learning (ML) have become considerably more important in recent years. However, in addition to many advantages, this has led to a sharp increase in resource consumption in the learning and operating phase of ML algorithms.

The integrated measuring instruments of chip manufacturers, on which the scientific community relies, only provide inadequate estimates of energy consumption. Measures based on these estimates could, in the worst case, lead to higher resource consumption.

This is where the "SAFES" project comes in, setting up a dedicated measurement laboratory in which the power consumption of the hardware can be precisely measured during the learning process and during operation. SAFES focuses on areas such as production and automotive and looks at industrial use cases (e.g. assembly robots, vehicle communication, autonomous driving).

Realistic energy models are then derived from the collected data and made available as open source frameworks. This should help the industry to design more sustainable AI systems in the future. In addition, a direct scientific and practical contribution will be made.