With artificial intelligence spreading ever more deeply into companies and everyday life, worries about its effects on the environment are growing. High-profile tech CEOs such as OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and previous Google CEO Eric Schmidt have openly mentioned how AI has to be employed to fight climate change. They propose that future uses of AI may be used to help develop cleaner energy technologies, enhance the ways carbon capture, and provide more sustainable food systems through innovations like lab-grown meat.

But most climate and tech specialists are still skeptical of such statements. According to them, such concepts are still highly speculative and are a distraction from what currently exists and works. Training and running generative AI models, such as language models like ChatGPT and Claude, are energy-hungry and computationally expensive operations. Not only do such models consume a lot of electricity to train but, more significantly, consume lots of electricity each time a person is using them, sustaining constant energy consumption.

Data centers that fuel AI are on their way to becoming the world's fastest-growing energy users. United States data centers alone consumed 3.7% of the nation's energy in 2023, McKinsey estimates, and could consume 11.7% by 2030. Data center energy consumption has grown 12% per year since 2017, the International Energy Agency reports, and will increase more than twice worldwide by 2030.

In spite of this increasing footprint, technology executives still largely believe in the long-term promise of AI for climate solutions. They believe that AI innovation will ultimately overcome the environmental price tag, allowing devices to optimize energy and accelerate breakthroughs for climate tech. This view has, however, been criticized for ignoring the urgency for near-term climate action and the presence of already-existence low-emission tech including solar panels, electric cars, and heat pumps.

Scientists caution that hope for imminent AI breakthroughs could cause delay on existing climate fixes. Accentuating AI as a climate fix poses the risk of overlooking more utilitarian and nearer-term action that can be taken now by governments and businesses. Political will to apply existing clean technologies is still a primary impediment, not absent high-tech creativity.

In addition, the distinction between lightweight energy-efficient AI software—e.g., generative AI for grid optimisation or climate modelling—and massive duty generative AI software is obfuscated in public debate. While AI may contribute to certain aspects of climate science, e.g., predictive modelling and simulations, its overall uptake through the utilization of generative AI contributes humongously to emissions without immediate environmental benefits.

Another point of concern is that despite pledges to reduce emissions, some of the companies leading AI development, including Microsoft and Google, have reported rising carbon footprints. As electricity demands rise due to AI and data centres, utility companies are delaying the closure of coal-fired power plants and planning new fossil fuel-based facilities. Although some stakeholders suggest nuclear energy could eventually offset AI’s growing energy needs, new nuclear infrastructure will take years to come online.

In conclusion, although AI has some capacity to augment climate science and operational effectiveness in certain places, present generative AI technology is not the solution to the climate catastrophe. Counting on AI for significant environmental intervention can result in more delay in addressing the pressing climate issue. Rather, investment in the roll-out of accessible clean energy technology and policy reaction should be made.

Source: San Francisco Examiner