Using AI sustainably - a social responsibility

Dr. Dina Barbian is Managing Director and sustainability expert at eco2050 Institut für Nachhaltigkeit - Institute for Sustainability GmbH. The institute was founded on the idea that we all have a social responsibility (CSR = Corporate Social Responsibility) towards future generations.

marketSTEEL: How did the eco2050 Institute come about and what expertise do you offer?

Dr. Barbian: The eco2050 Institut für Nachhaltigkeit - Institute for Sustainability GmbH in Nuremberg was founded as a spin-off of the University of Erlangen-Nuremberg in 2012. It supports companies in the development of sustainability strategies. The institute was created to establish sustainability as a cross-cutting topic in all industries and disciplines. To this end, research on the topics of sustainable digitalization and sustainable technologization was carried out at a very early stage, which is also reflected in our publications.

Its key areas of expertise include writing sustainability reports and conducting climate assessments. The eco2050 Institute conducts research on sustainable ICT systems, sustainable AI, the circular economy and green and orange hydrogen. It offers seminars and events on many different interdisciplinary topics such as greenhouse gas accounting, circular economy, decarbonization, etc.

marketSTEEL: What motivates you - what is your vision?

Dr. Barbian: Our vision is a world that fully implements the 17 Sustainable Development Goals of the United Nations from 2015. The 17 Sustainable Development Goals (SDGs) are the most comprehensive framework to date for implementing sustainability for individual countries around the world. The aim is to implement the 17 goals with their 169 sub-goals by 2030. This is why this global action plan is also called “Agenda 2030”. There is a slogan: “Leave no one behind!” (Leave no one behind!)

At our institute, we try to take the 17 objectives into account in all our projects as far as possible.

marketSTEEL: You work closely with the ESCADE project. What is ESCADE's task?

Dr. Barbian: ESCADE stands for Energy-Efficient Large-Scale Artificial Intelligence for Sustainable Data Centers. The aim is to develop sustainable solutions that ensure operation in energy-efficient data centers with innovative chip technologies by designing use cases and developing AI applications based on cognitive AI technologies.

Fig. 1: ESCADE project at a glance

The eco2050 Institute for Sustainability GmbH is involved in ESCADE as a subcontractor for the DFKI (German Research Center for Artificial Intelligence). Our task is to make the sustainability of AI data centers measurable. To this end, we have developed an ESG indicator model to measure the sustainability of data centers. This will be presented at the Hannover Messe in early April 2025. ESCADE was selected for the stand of the Federal Ministry of Economics and Climate Protection (BMWK) at the Hannover Messe.

The sustainable, resource-efficient design of data centers and AI systems also contributes to the achievement of UN Sustainable Development Goals 9, 12 and 13. In addition to the ESG indicator model, we are currently developing another framework model for measuring the sustainability of AI systems. This will make it possible to decide which AI system is more sustainable than another. The concepts will then be tested and evaluated in the use cases “Sustainable steel industry through resource-efficient AI” and “Large, resource-efficient NLP models”.

marketSTEEL: You are working together with partners. Who are they?

Dr. Barbian: The partners are NT Neue Technologie AG, Stahl-Holding-Saar GmbH & Co. KGaA, SEITEC GmbH, the Technical University of Dresden, the University of Bielefeld and the Austrian partner Salzburg Research Forschungsgesellschaft. The consortium is led by DFKI under the direction of Prof. Dr. Wolfgang Maaß.

Fig. 2: Criteria for evaluating the sustainability of AI

marketSTEEL: You talk about ESG indicators for sustainable data centers. What exactly do you have in mind?

Dr. Barbian: One of the central tasks in the ESCADE research project (see Fig. 1) is the sustainability assessment of AI throughout its entire value chain using a key performance indicator system. Using two use cases that represent different AI tasks, novel AI hardware (so-called neuromorphic hardware, whose functionality is inspired by nerve cells) and AI algorithms adapted to it are being tested. Initial preliminary tests in laboratory experiments have shown that the new hardware and improved algorithms consume up to 80% less energy than conventional technology, depending on the task.

With the help of the developed ESG indicator system, it will be possible to evaluate the sustainability of AI systems and thus make transparent and informed decisions. The term “ESG” comes from the world of finance and describes the threefold division of sustainability into environmental, social and economic aspects. The three sustainability-related areas of corporate responsibility are: “E” stands for Environment. Social (“S”) includes societal aspects and governance (“G”) refers to sustainable corporate management.

Another important aspect is the transfer of research results into practice in order to help shape the transformation through AI for greater sustainability. There is already initial potential for developing sustainable AI business models.

ESCADE's explicit objective is to make the sustainability aspects (e.g. resource consumption or greenhouse gas emissions per service unit) of AI tasks transparent in addition to the financial parameters (e.g. costs per service unit), thus enabling fact-based and well-founded decisions to be made on the basis of an ESG indicator model.

Based on the relevant standards, norms, regulations and laws, 107 key figures were compiled as part of the project, which were sorted according to the categories Environment, Social and Governance (ESG). Figure 2 shows the aspects considered in each case in the form of a mind map.

More information:

https://eco2050.de/

https://escade-project.de/#Konsortium

Copyright: © eco2050 Institute for Sustainability