1.2 From Chips to Thoughts: Building Physical Intelligence into Robotic Systems

Abstract

The rapid growth of AI technologies has brought unprecedented advancements across numerous domains, from healthcare to autonomous systems, yet this progress has been accompanied by substantial energy demands. Generative AI has revolutionized the accessibility of advanced machine learning models, putting powerful AI capabilities directly into the hands of everyday users. Tools that generate text, images, video, and audio have made AI more democratic, fueling creativity, productivity, and innovation across industries. This explosion of generative AI applications, from chatbots to art generators, has transformed AI from a niche tool into an indispensable asset in our digital lives, as many can now access these technologies through their phones or computers. However, this widespread accessibility comes with significant trade-offs, particularly in terms of the energy required to train and deploy these enormous models. Large generative models such as GPT, DALL-E, and other multi-modal AI systems demand immense computational resources, translating to high energy consumption both during the extensive training process and when they are used in practice.