Photo by anouar olh on Pexels

As artificial intelligence models become more efficient and less expensive to run, a counterintuitive economic principle, the Jevons Paradox, is raising concerns. The paradox suggests that increased efficiency in resource use can actually lead to higher overall consumption of that resource. In the context of AI, this means that lower costs per inference or training run might encourage significantly increased usage, potentially resulting in a net increase in spending on AI infrastructure and services. A discussion on this topic has emerged within the AI community, sparked by a recent Reddit post [https://old.reddit.com/r/artificial/comments/1mri8jf/jevons_paradox/]. The community is now exploring strategies for effectively tracking, managing, and ultimately controlling AI-related expenses in the face of potentially escalating demand.