Global AI Initiative Database

Systematic survey and target group-specific preparation of the current state of research state of research on ecologically sustainable AI systems and AI applications and development of a guideline on Green AI.

Green AI is a highly relevant topic and greater attention should be paid to green AI when discussing the requirements for AI applications. In view of the ever-larger language models and their roll-out to universally accessible applications (e.g., integration into search engines) as well as their spread to even more computationally intensive modalities (e.g., generation of video content), it is to be expected that the energy needs for training and inferencing AI systems will increase sharply, and that we are currently only at the beginning of this development.There is already an understanding of how AI applications can be made more sustainable (renewable energy, less computationally intensive algorithms, new architectures such as edge AI or spiking neural networks), but many research questions are still open and to date remain underappreciated among the target group of AI developers and policy makers.The aim of this R&D service is therefore to systematically record the current state of research on sustainable AI systems and applications for specific target groups. This know-how will be used to draw up a guideline for implementing sustainable AI applications. This guideline will be made available to all projects of FFG calls relevant to AI, and to the ministries responsible for AI applications.Objectives and impact:- Answering or identifying open research questions in relation to the sustainability to the sustainability of AI- Dissemination of knowledge on how AI can be designed and used sustainably and, as a result, reducing the consumption of resources and emissions through AI systems

Federal Ministry for Climate Action, Environment, Energy, Mobility, Innovation and Technology (BMK)