Abstract: Artificial intelligence (AI) holds enormous potential for enabling scientific breakthroughs, accelerating business growth, and tackling global challenges. Nonetheless, the evergrowing demand for AI has created a huge environmental footprint, which, if not properly addressed, can potentially exacerbate climate change. For example, even putting aside the environmental toll of chip manufacturing, training a large generative model like GPT-3 can easily consume hundreds of megawatt-hours of electricity, generate tonnes of carbon emissions, and evaporate millions of liters of fresh water for cooling. Simultaneously, AI’s environmental inequity — the environmental cost of AI disproportionately impacts certain (often marginalized) communities — has also been emerging as a worrisome outcome and increasingly recognized as a roadblock to responsible AI.
In this talk, I will present the challenges as well as my research towards environmentally responsible AI. I will begin with algorithmic foundations for managing AI systems and present provably robust online algorithms that embed learning-based policies into competitive designs. Next, I will show our work on learning-augmented scalable neural architecture designs for sustainable AI on the edge. Finally, I will discuss our work that leverages AI to enable sustainability and equity in broader societal cyber-physical infrastructures.
Bio: Shaolei Ren is an Associate Professor of electrical and computer engineering at the University of California, Riverside, where he is also a cooperating faculty member in the computer science and engineering department. Broadly focusing on AI + sustainability, he is interested in tackling algorithmic and systems challenges to build a sustainable and equitable future. His research addresses water footprint and environmental inequity in AI/computing systems, and has generated broader societal impacts, including fostering public awareness of environmentally responsible AI through worldwide coverage in ~100 countries (e.g., The Associated Press, The Wall Street Journal, and Nature Briefing) and informing technology policies on future AI. He received the NSF CAREER Award in 2015 and multiple Best Paper awards (including ACM e-Energy’16 and IEEE ICC’16). He holds a Ph.D. degree from the University of California, Los Angeles.