Tech companies are blending traditional AI with generative AI when they claim that this energy-intensive technology can help combat climate change, a recent report suggests. The findings highlight that most statements about AI’s potential to reduce climate impact refer to conventional machine learning, not the energy-demanding chatbots and image generators driving the industry’s growth.
The analysis of 154 statements, commissioned by nonprofits like Beyond Fossil Fuels and Climate Action Against Disinformation, found no evidence that popular tools like Google’s Gemini or Microsoft’s Copilot have led to any significant reductions in greenhouse gas emissions.
Ketan Joshi, an energy analyst and author of the report, criticized these tactics as “greenwashing,” which mislead the public about actual environmental benefits. He compared this approach to how fossil fuel companies exaggerate their investments in solar energy while understating their carbon footprints.
The report’s claims were mainly drawn from an International Energy Agency (IEA) analysis. This analysis highlighted a mixture of academic publications, corporate claims, and unsupported assertions. Many claims from Google and Microsoft lacked solid evidence.
The discussion around technology’s climate benefits often lumps different types of AI together. Sasha Luccioni, AI and climate lead at Hugging Face, noted that not all AI is created equal. She pointed out that generative AI and large language models have a larger negative impact on the environment compared to older, traditional models that are often more beneficial.
A staggering 74% of claims about AI’s environmental impact were found to rest on shaky evidence or none at all. For instance, a widely cited claim that AI could reduce emissions significantly by 2030 emerged from a blog post with little substantiation.
The reality is that data centers, which house AI infrastructure, are becoming increasingly energy-intensive. They account for 1% of global electricity consumption, but this figure could double in the U.S. by 2035. The IEA warns they could drive up to 20% of electricity demand growth in wealthier countries by the decade’s end.
While a simple query to tools like ChatGPT might use as much energy as a light bulb running for a minute, more complex tasks can significantly increase power needs. Industry experts are raising alarms about these growing energy demands.
Google stated its emissions reduction claims are backed by a strong scientific foundation. Microsoft did not respond for clarification, and the IEA has remained silent on the matter. Joshi emphasized the importance of separating big issues from small solutions in discussions about climate.
In conclusion, as we move forward, it’s crucial to accurately understand the impact of AI on our planet. Misleading information can detract from the pressing issues and real solutions needed to address climate change effectively.
