Reducing the Carbon Impact of Generative AI Inference (today and in 2035)
Andrew A. Chien, Liuzixuan Lin, Hai Thanh Nguyen, Varsha Rao, Tristan Sharma, Rajini Wijayawardana
Abstract
Generative AI, exemplified in ChatGPT, Dall-E 2, and Stable Diffusion, are exciting new applications consuming growing quantities of computing. We study the compute, energy, and carbon impacts of generative AI inference. Using ChatGPT as an exemplar, we create a workload model and compare request direction approaches (Local, Balance, CarbonMin), assessing their power use and carbon impacts.
Topics & Concepts
InferenceGenerative grammarComputer scienceGenerative modelWorkloadCarbon fibersArtificial intelligenceMachine learningAlgorithmOperating systemComposite numberFerroelectric and Negative Capacitance DevicesAge of Information OptimizationExplainable Artificial Intelligence (XAI)