AI, Energy, and Water скачать в хорошем качестве

AI, Energy, and Water

Bert Baumgaertner (UI) and Joel Gladd (CWI) presented on AI's energy and water footprint. We'd promised jobs too, but the environmental research alone filled the session. Jobs and the workforce will get its own webinar. Bert opened with interpretive framing around individual vs. collective obligations and descriptive vs. normative claims. He also provided guidance on what works or doesn’t when using arguments by analogy. This framing resurfaced throughout the presentation. From there, we looked at what data centers actually do, how efficiency is changing as legacy systems transition to newer closed-loop designs, and why the energy source matters as much as the cooling technology. Then we got into the cost of prompting. "AI uses x energy" is a meaningless statement without specifying the type of prompt. The compute required for a simple text query vs. a reasoning model vs. video generation varies by orders of magnitude. We also saw how, for many faculty, Zoom meetings dominate their energy and water footprint. We compared a variety of Gen Z vs. faculty scenarios to see how daily digital habits could be translated into water and energy calculations. To help with in-class workshops and discussions, Joel created a companion website (see below) for students and faculty to inventory their daily digital habits and see the water and energy usage totals. The idea is to navigate these conversations with students by understanding how it’s part of our entire digital lifestyle.