Big Tech Says Generative AI Will Save the Planet. It Doesn't Offer Much Proof (www.wired.com)

In late 2023, Google claimed that generative AI could reduce global greenhouse gas emissions by 5-10% by 2030, a statement that quickly gained traction in the media and academic circles. However, energy researcher Ketan Joshi found that the basis for this estimate was weak, originating from a prior analysis by BCG that relied on vague client experiences rather than robust research. Furthermore, Google later acknowledged in its sustainability report that the infrastructure needed to support AI technologies was actually increasing its own emissions, undermining the original claim. This inconsistency raises serious questions about the validity of the assertions made by leading tech companies regarding AI's potential environmental benefits. Joshi's recent report scrutinizes over 150 AI-related climate claims, revealing that only a quarter were backed by credible academic research, with many lacking evidence altogether. The findings suggest a pattern of conflating traditional machine learning, which is less energy-intensive, with consumer-focused generative AI, which relies on vast computational resources. Experts argue that the narrative pushed by big tech centers around large-scale models obscures the viability of smaller, more efficient AI solutions that could provide climate benefits without the significant environmental costs. The ongoing lack of transparency regarding the energy demands of AI systems creates challenges in accurately assessing their climate impact, prompting calls for greater disclosure from tech companies about their energy usage associated with AI development.