Eco-Friendly AI: A Guide to Energy-Efficient Techniques Across the AI Life-Cycle

As AI’s influence expands, numerous concerns about potential risks emerge, ranging from job displacement and lack of transparency to cybersecurity threats. However, the urgency of climate change—a critical challenge impacting ecosystems and livelihoods—emphasizes the need to address AI’s energy usage and carbon emissions. In recent years, researchers have developed techniques to mitigate the environmental footprint of AI. This work presents an overview of these recent techniques aimed at making AI more sustainable throughout its life-cycle. It also compares these techniques by their impact, showcasing their effects in a Cheat-Sheet for practical use.