Major Tech Claims Generative AI Will Rescue the Planet, Yet Lacks Substantial Evidence

However, many of these assertions, it appears, have minimal—if any—real evidence to support them.

Joshi has authored a new report, released on Monday with backing from various environmental organizations, which seeks to evaluate some of the most notable claims regarding AI’s potential to benefit the planet. The report examines over 150 assertions made by tech firms, energy groups, and others concerning how “AI will contribute positively to climate change.” Joshi’s findings reveal that only a quarter of these claims were substantiated by academic studies, while more than a third lacked any public evidence at all.

“Individuals make statements about the societal effects of AI and its impact on the energy system—those statements often lack proper substantiation,” notes Jon Koomey, an energy and technology researcher not involved in Joshi’s report. “It’s crucial not to accept self-serving claims without skepticism. While some may hold true, caution is necessary. Many individuals make these declarations with little backing.”

Another significant aspect the report addresses is the specific form of AI to which tech companies are referring when they suggest AI will aid environmental efforts. Various AI types demand less energy than the generative, consumer-oriented models that have garnered attention lately, which necessitate substantial compute power—and energy—for training and operation. Machine learning has been widely employed across multiple scientific fields for decades. However, it’s the large-scale generative AI—particularly tools like ChatGPT, Claude, and Google Gemini—that are the focal point of much infrastructure expansion by tech companies. Joshi’s research indicates that nearly all claims examined blurred the lines between traditional, less energy-demanding AI and the consumer-targeted generative AI driving substantial data center development.

David Rolnick, an assistant professor of computer science at McGill University and chair of Climate Change AI, a nonprofit promoting the use of machine learning for climate solutions, expresses a different concern than Joshi regarding the origins of Big Tech’s information on AI’s climate impact, citing the challenges of quantitatively proving such effects. Nevertheless, Rolnick believes the distinction between the types of AI companies present as vital is an essential part of this discussion.

“My issue with the claims made by large tech companies about AI and climate change isn’t that they’re not completely quantified, but rather that they often lean on speculative AI that doesn’t currently exist,” he states. “The level of conjecture surrounding the potential future uses of generative AI is alarming.”

Rolnick highlights that from improving grid efficiency to developing models that assist in discovering new species, deep learning is actively employed in numerous sectors globally, already contributing to emissions reduction and combating climate change. “This is distinct from claims like ‘At some future point, this may be useful,’” he explains. Furthermore, “there exists a disconnect between the technologies that big tech companies are developing and the ones that are genuinely delivering the results they claim to endorse.” Some firms may promote algorithms that, for instance, enhance flood detection, using these as examples of beneficial AI to market their large language models—despite the fact that the algorithms aiding in flood prediction differ significantly from a consumer-oriented chatbot.