"Switch language"

Principles for Green Artificial Intelligence

1 min read

Artificial intelligence (AI) can help protect the climate. At the same time, the technology already consumes substantial energy and resources. That makes it important to promote the technology's sustainable development and use. Our nine principles for green AI help guide us. 

Infographic showing DT's green AI principles
Deutsche Telekom's nine principles for green AI. © Deutsche Telekom

The principles provide guidance on how AI solutions can be developed and deployed in a more environmentally sustainable way. They also show how risks - such as a significantly growing CO2 footprint - can be addressed early on. We want to make the development and use of AI within Telekom more sustainable, while also inspiring others: companies and institutions, as well as policymakers and researchers. The goal is to align AI with sustainability from the very beginning.

 

1. Green electricity

Infographic showing DT's 1st priciple for green electricity
AI applications require more electricity than conventional IT applications. That is why we rely on renewable energy throughout the value chain and encourage our partners to do the same. © Deutsche Telekom

 

2. Reusing resources

Infographic showing DT's 2nd green principle on reusability in the value chain
We reuse hardware, software, AI models, and data throughout our value chain. This helps us work more flexibly and efficiently while avoiding unnecessary energy use. © Deutsche Telekom

 

3. Transparent CO2 footprint

Infographic showing DT's 3rd green principle on transparent CO2 Footprint
Our AI development teams consider the CO2 emissions from hardware and software and assess how changes to hardware and software affect the environmental footprint. © Deutsche Telekom

 

4. Dynamic scaling

Infographic showing DT's 4th principle on dynamic sizing
Oversized hardware consumes unnecessary energy. We align our IT resources with demand. And we intelligently scale down what we do not need for modeling, training, or operations. © Deutsche Telekom

 

5. Optimized AI models

Infographic showing DT's 5th green principle on Optimized AI Models
We select optimized, tested AI architectures and models for specific use cases. Where the task allows, we follow a modular approach and reuse software components. This helps us work more efficiently and reduce energy consumption. © Deutsche Telekom

 

6. No duplicate work

Infographic showing DT's 6th green principle on no duplications
We do not reinvent the wheel every time we develop software for new AI applications. For similar use cases, we avoid duplicate work and leverage synergies. That is why we create transparency and reuse code whenever possible. © Deutsche Telekom

 

7. Green coding

Infographic showing DT's 7th green principle on green coding
Programming languages differ in their energy consumption. Because we want to develop with energy awareness, we code as efficiently as possible. © Deutsche Telekom

 

8. Keep it as simple as possible

Infographic showing DT's 8th green principle on simplicity
We choose AI models that are as simple as possible, as long as they can handle the specific task. Once we have selected the right AI model, we focus on algorithm efficiency to use fewer resources. © Deutsche Telekom

 

9. End-to-end responsibility

Infographic showing DT's 9th green principle on end-to-end responsibility
We take responsibility for the entire value chain. That is why we regularly review the carbon footprint of the hardware and software used in our AI applications. We are committed to end-to-end responsibility for green AI. One focus is generative AI. © Deutsche Telekom

More on this topic