Do the benefits of AI data centres outweigh the environmental costs?
Artificial Intelligence data centres are expanding across the UK and around the world. Tech giants and governments see these as essential infrastructure for future prosperity: the engines of economic growth, scientific innovation, improved healthcare and national security.
Critics, however, point to the environmental costs. These vast warehouses of IT equipment eat up huge amounts of power; some require millions of litres of water to keep cool. Do the promised rewards justify the sacrifices?
Is society being swept along by an AI ‘arms race’ we didn’t consent to? Are we right to put our faith and future in the hands of tech billionaires or are we simply enabling more consumption?
On BBC Radio 4’s the Moral Maze, experts in the fields of charity, tech and philosophy offer up their own ideas around these competing visions of the future.
We need greater transparency about energy use
Owen Espley – campaign manager at the environmental charity Global Action Plan – argues that data centres are “gobbling up” our energy.
...we need a framework in place to make sure that it's done sensibly, within environmental limits.
“The amount of energy we have available is limited, the amount of water that's available is limited,” he stresses.
He believes there should be a pause on building more data centres until a coherent plan is in place.
“The risks of causing catastrophic runaway climate change are clear,” he argues. “We can move forward with this technology, but we need a framework in place to make sure that it's done sensibly, within environmental limits.”
Greater transparency is also important. When we impose harms on communities, they have a right to scrutinise the case for any particular development, he argues.
“We're not getting transparency from industry about their energy use, their water use,” he states.
Data centres are infrastructure, not extravagance
Michael Tobin is a tech entrepreneur who builds data centres, and he takes a very different stance.
“You've got to think about data centres as a piece of infrastructure,” Michael says. “It's utility.”
“The data centre doesn't consume anything. We do.”
Many of us will have hundreds, to tens of thousands, of pictures from our phones “whirring away” in a data centre somewhere. The data centre is merely an efficient way of delivering the resources that we, as a society, demand.
“When you put compute power into one space and you run it professionally, you get the maximum power usage ratio,” he argues.
He also suggests that around only 2% of our total energy consumption in the UK is data centre power consumption.
“That's about half of what electric cars are consuming.”
AI is offering tangible benefits that we can’t ignore
“There's something very exciting about AI,” states feminist author Ella Whelan. “To have knowledge at your fingertips.”
[AI] is a democratising tool... It's changing people's quality of life.
People who may have previously found themselves ignored are now using AI to write impressive legal complaints or letters to their council – and getting their cases heard.
“This is a democratising tool,” says the writer. “It's changing people's quality of life.”
Michael references his own medical case. During a recent colonoscopy, it was AI rather than the “expert doctor” who identified three polyps – potentially stopping cancer in its tracks.
Many argue that the best way for us to find a solution to the climate crisis is to allow AI to help us.
Although Owen from Global Action Plan argues that, when it comes to solving climate change, the evidence just isn't there.
“We're also seeing lots of claims made about the use of AI in war to accelerate destruction and human suffering,” says the campaigner.
Andrew Copson, Chief Executive of Humanists UK, remains optimistic about its uses.
“Already viruses have been modelled and diseases have begun to be profiled in ways they couldn't be if we weren't using these models and so the benefits might be uncertain, but they can't count for nothing,” he states.
Could AI be managed in a more democratic way?
“Technologies don't have to be owned by a small number of companies that have amassed enormous amounts of wealth,” suggests Owen. “They can be used and democratised so that we actually have control over it.”
Dr. Carissa Véliz, Associate Professor of Philosophy at the Institute for Ethics in AI at Oxford University, agrees.
“We can design it to support democracy and to be more environmentally friendly and more reliable,” she says. “It’s not anti-technology, it's better technology and more democratic governance.”
The billionaire tech moguls have proven themselves to be untrustworthy, designing products that are only for profit and not, for example, truth tracking, argues the academic.
“We should be investing in European companies that are doing things different. And part of what they're doing different is that they're not building data centres. They're building small data servers for people to have control of their own data.”
For Carissa, it’s about making AI models and management better, not banning it.
It’s not all or nothing – progress can be staged and reversible
“Technology is like toothpaste. Once you squeeze the tube, you can't put it back,” argues Michael.
Maybe innovation is inevitable directionally, but the pace of it is definitely not inevitable.
But Dr Carlo Codasco, a moral philosopher at Manchester University who focuses on the ethics of AI, doesn’t buy this idea of irreversibility.
“Maybe innovation is inevitable directionally, but the pace of it is definitely not inevitable,” he says.
He argues that we can learn about the benefits and costs of AI as we go along, and simply make sure that we don't commit too much (or too little).
“The best thing we can do is to preserve some kind of optionality and flexibility,” he states. We do this by making investment choices that don't leave us with only bad options.
“We're going to be able to make decisions at later stages as we go by and we learn about things.”
“Let's just do things in a staged way.”
He’s also optimistic about the environmental impact.
“We have a lot of closed-loop data centres now,” he says. This means the water for cooling is reused. “Bottlenecks in terms of energy – I think these are solvable one way or another.”
We need to weigh risk against regret
We have a responsibility to future generations, argues Mona Siddiqui, Professor of Religion and Society at King's College London.
“What gives us the right to make irreversible decisions about our environments on behalf of people who didn't ask for it? And who's going to be accountable if it doesn't work?” she asks.
There’s another way of posing the question, suggests Carlo: “What are future generations going to say if we slow down the progress of AI?”
Do we owe it to future generations not to press pause? Is it about effective shaping through government regulation, not refusal?
“There has to be some kind of innovation,” argues Ella. “There has to be some kind of change. So we should row about what change there is. But simply saying there can be no change, everything must stop, that isn't satisfying to anyone.”
Even if we did decide to stop, others would carry on, stresses Andrew.
“Are we willing to let authoritarians and possibly those that you might think of as the worst governments in the world move on and progress with this technology rather than try to set the terms of that ourselves?”
“That's what might matter more to our future generations.”
