To be Artificial Intelligence for sustainability or not to be sustainable Artificial Intelligence

Sustainable electrical energy (SEE) represents a direct and indirect critical enabling factor for sustainable development. Artificial Intelligence (AI) has emerged as a disruptive technology that accelerates the transition toward full SEE, but concurrently, presents a fatal issue. This study inspects the benefits and limitations of AI applications in the context of SEE. The primary finding highlights a common assumption: AI utilization in areas such as green power generation, electric vehicles, and so on is often deemed, per se, sufficient to achieve full SEE. However, the study highlights a critical shortcoming of this perspective: the absence of a holistic SEE. Achieving full SEE requires integrating the life cycle assessment, which evaluates the environmental impact from raw material extraction to end-of-life management. This ensures considering the environmental adverse effects of green technologies such as renewables and electric transportation, thus embracing a real technology neutrality. Additionally, a thorough consideration of electrical energy production and consumption is necessary. Moreover, it emerged that AI-based holistic planning enabling a fully green-supplied power system has not been sufficiently investigated so far. In conclusion, the first part of the study has brought out that a transition toward fully sustainable electricity imposes that AI considers the design for sustainability paradigm and a holistic view of sustainability that combines life cycle assessment and exploitable electrical energy. The large room for improvement in the adoption of AI for full SEE and its imperative priority ask for an urgent research effort of academia and industry. With this in mind, some figures of merit have been discussed. The second part of this study investigates the main sustainability challenges of AI. The analysis of the (e-)waste, pollution, and energy demand has highlighted that AI widespread use is untenable with the current technologies. The key issue is the unsustainable growth of electrical energy consumption due to AI, which incontrovertibly emerges as the core challenge, because the research effort has focused on achieving even-increasing accuracy regardless of the energy consumption. Large diffusion of humanoids that would exploit various AI tools to face different problems while using large language models and generative AI systems will skyrocket energy demand. Natural evolution has inherently optimized the human brain for energy efficiency, while the evolution of AI has led, conversely, to energyinefficient outcomes. Therefore, a worldwide research effort must be lavished on developing low-energydemand AI while keeping sufficient accuracy. In conclusion, all the research efforts have focused so far on providing AI with human skills until reaching super-human abilities, overlooking the crucial one: very lowenergy-demand high-computing aptitude.