Every time you use ChatGPT, half a litre of water goes to waste

Tech giants including OpenAI, Meta and Google are being urged to tackle AI’s dirty secret – its massive power problem.

Generative AI already uses as much energy as a small country and is predicted to rival that of Japan within a year. Such searches use 10 times the energy of a normal web search, and the technology has tripled the energy requirements of the entire tech sector in just two years.

Speaking before a lecture this week about the environmental, political and social impacts of generative AI at Victoria’s State Library, Australian AI expert Professor Kate Crawford said the technology’s rampant popularity was worsening what was an already dire climate crisis.

Professor Kate Crawford says AI’s rampant popularity is worsening the climate crisis. Credit:

Crawford, who was recently named on Time magazine’s list of the 100 most influential people in AI, warns that AI systems are already reshaping the planet in lasting and often hidden ways.

“We are under enormous pressure to try to decarbonise by mid-century, and if we don’t, we’re looking at the sort of spiralling planetary impacts that will affect all of us,” she said.

“You can’t outrun climate change. So even for tech billionaires, this should be an issue that we should all be able to get behind because, frankly, we just don’t have the time to waste to be building systems that are using as much energy and creating carbon footprints as big as entire industrial nations.

“This is not the time to be doing that.”

It’s not just ballooning energy consumption causing concern. The power-hungry data centres underpinning the AI systems require vast amounts of fresh water for cooling, said Crawford, much of which is simply evaporated.

She said each interaction with ChatGPT was the equivalent of pouring half a litre of water on the ground.

“We’re wasting a lot of water with these systems, and very few people realise that it’s a major problem,” she said. “That’s why I think personally the No.1 priority for the sector should be sustainability. Not the AI race.”

Earlier this year, OpenAI chief executive Sam Altman said an energy breakthrough would be necessary for future artificial intelligence.

OpenAI chief executive Sam Altman has said an energy breakthrough will be necessary for future artificial intellgence.Credit: Bloomberg

“There’s no way to get there without a breakthrough,” he said. “It motivates us to go invest more in fusion.”

Altman has personally invested in nuclear fusion company Helion Energy, which has signed a deal to provide nuclear energy to Microsoft.

Crawford views nuclear as a “quick fix” for a larger, longer-term issue. For Crawford, there are a few ways to tackle it. One is for researchers to change the way they build AI systems, in a way that is less energy intensive, prioritises sustainability, and relies on renewable energy and recycled water to reduce the impact on the planet.

Another is for tech giants to treat the sustainability of a generative AI model as importantly as its capabilities or its marketing.

Australia, she believes, has a critical role to play. As a country already grappling with climate impacts, its policies could set global precedents, she said.

Crawford also pointed to last week’s Senate select committee report that emphasised a need for sustainability and transparency from the tech giants. One of its recommendations was for the federal government to “take a co-ordinated, holistic approach to managing the growth of AI infrastructure in Australia to ensure that growth is sustainable, delivers value for Australians and is in the national interest”.

Kyndryl executive and former Tesla executive Faith Taylor.

Crawford envisions a future in which AI contributes to solving the climate crisis, rather than exacerbating it.

“What we would love to see is artificial intelligence that can help us solve the climate crisis without contributing to it at the same time.”

That’s also the view of Faith Taylor, the executive in charge of sustainability for Kyndryl, the world’s largest IT services provider.

She was previously global environmental social governance leader for Tesla, working with Elon Musk.

“I like to say ‘sustainability is a team sport’,” Taylor said. “There is an urgent need to address climate change, and it takes collective action ... Whether you’re a cloud provider or an AI developer, each party plays a role in minimising our collective environmental impacts as they operate within a broader system that includes governments, energy providers and hardware manufacturers.”

Taylor said AI could help accelerate the transition to renewable energy by forecasting weather patterns, monitoring supply chain performance, and even by making itself more energy efficient over time as companies develop low-energy AI models.

Crawford said that after two years of “shock and awe” with the blockbuster launch of ChatGPT and its subsequent adoption, we’re now at a period of reflection with the technology.

“I do think we’re at a turning point,” she said. “We’ve reached a point now where we can look back and say, ‘OK, what’s useful, what is actually working for us and what isn’t working for us?’

“We want to make sure AI is actually really contributing to society in all of the right ways, rather than causing problems that we’ll have to solve later.”

For Future Reference 2024: Mapping Planetary AI with Professor Kate Crawford will be held at 7.30pm on Thursday, December 5 in the Conversation Quarter, State Library Victoria. For Future Reference is an annual lecture series endowed by the Helen McPherson Smith Trust.

The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to get it every weekday morning.