Sustainable Use of Generative AI

The considered, careful and critical use of generative AI also involves reflecting on whether other (technical) support options are associated with a lower resource consumption. Generative AI systems consume significant computational and energy resources, particularly during the training of large models and the operation of large-scale data centers.

A simple internet search or the use of conventional software tools can often deliver similar results in a comparable amount of time while using significantly less resources.

The following aspects can also contribute to a more sustainable approach:

• Needs-based use: Only use AI when there is added value, limiting the scope of queries (e.g. no unnecessarily long prompts or multiple calculation runs with no gain in content).
• Conscious choice of tools: Give preference to platforms that commit to transparency in terms of energy consumption, CO₂ footprint and water use or use sustainable Computer Centers (e.g. with renewable energy).
• Local vs. cloud solutions: Where possible, run smaller language or analytics tools locally (e.g. offline writing and proofreading tools, local analytics or statistics software) rather than using large cloud models for simple tasks.
• Resource-saving prompting: Formulate precise questions to avoid repetition and unnecessary computing work.
• Long-term perspective: Check whether results or AI-supported workflows can be archived in order to avoid multiple queries and strengthen knowledge transfer and gain.

Reference is also made to the sustainability strategy of the University of Regensburg (external link, opens in a new window), which emphasizes the responsible use of digital technologies. In addition, the AI Campus course AI and Sustainable Development Goals (external link, opens in a new window) offers practice-orientated foundations for a reflective and resource-conserving approach to AI.