Digital Transition and Artificial Intelligence for Good: Towards Ethical and Sustainable Energy Consumption?

The digital transition has become ubiquitous, accelerating markedly since many years. Although countries and companies are 'dematerializing' many activities, this shift remains deeply material, manifesting discreetly in data centers, network infrastructures, and the production of diverse digital devices. Among all digital technologies, artificial intelligence (AI) has shown recently a significant impact on energy consumption, primarily due to the computational intensity of training AI models and data processing. On the other hand, AI can also help reduce the energy consumption and emissions by optimizing the energy systems and enabling the transition to low-carbon sources. But this could also provoque a rebound effect as observed with other digital technologies. Being aware of the Jevons' paradox, this article synthesizes findings from various studies to highlight two key points: first, the growing environmental footprint of digital technologies, particularly in terms of electricity consumption; second, the necessity and opportunity of steering digital development towards greater sustainability. We argue that a more frugal digital model is not only feasible but also essential to support the broader ecological transition, notably with the massive adoption of artificial intelligence. The reducing of the energy footprint of artificial intelligence involves a combination of efficient model design, hardware improvements, data management, and policy measures to promote a sustainable development and deployment. These efforts are crucial to harness the benefits of AI while mitigating its environmental impact.