AI supercomputers run on fossil fuels, water, even nuclear energy: How hard is it hitting our planet?

“Summarise this text for me.”

“Make me an image of Mars, but goofy.”

“Can birds think?”

These are some of the useful and outlandish requests that we frequently make to AI software like ChatGPT, Dall E and Bard.

Some of us even talk to chatbots like Meta AI just for fun or to kill time.

But somewhere in a data centre in a remote part of the world, all our requests are being processed. The answers presented on our screens, to be hurriedly consumed, do not appear out of thin air. They are processed using massive computing power, often on what are called ‘supercomputers.’

These supercomputers are run on real-life resources like water, fossil fuels, and even nuclear energy.

Sasha Luccioni, the climate lead for the AI company Hugging Face, and her team conducted researchand foundthat 500 metric tonnes of carbon dioxidewasproduced just in the training of ChatGPT’s GPT3 model—the equivalent of over a million miles driven by average gasoline-powered cars.

So why are we not focusing on artificial intelligence’s not-so-hidden cost to the environment? How does this fare in the era of the climate crisis?

The Data Centre: A massive fuel-eating system

Contrary to popular belief, each colourful WhatsApp sticker that we ask Meta AI to generate is processed at a physical data centre.

A data centre is a large group of linked computers used for remote storage and processing.

In a nutshell, it processes our requests and gives us the final products.

But these centres consume massive amounts of energy and use natural resources.

They run on fossil fuels, which stress an already overused planet.

The Cost of Cooling

All machines, including massive groups of computers, produce a lot of heat while working.

Just like an air conditioner cools the room while producing heat, a processor or a computer also produces heat while computing.

This heat needs to be cooled down.

To do this, water, often fresh water, is used by companies. Traditional air conditioning systems, which use less water but more electricity, are also used to reduce the temperature.

This creates further pressure on the environment where the data centres are located, causing local communities to face acute water shortages.

After a lengthy legal battle, it was revealed that Google’s data centres in the city of Dalles, Oregon, used a quarter of the town’s water supply.

The environmental costs of AI are high, but can humanity function without this advanced technology?

Necessary Evil?

Artificial intelligence has done wonders in battling the climate emergency, to an extent. Using AI models, scientists can now create detailed maps of flood-affected areas, identify at-risk communities, and help authorities take action.

Meteorological departments across the globe depend on advanced AI to predict the weather, making such technology crucial for the world's food system.

“AI useful to aid in the prediction of which areas are identifiable as most vulnerable to heat waves and flash floods, but is not likely to support any meaningful mitigation measures” says Maulik, a climate activist.

Artificial intelligence is now essential to human well-being, but the conscious and sustainable use of AI is the need of the hour if we are to continue reaping its benefits.

Sustainable AI

Considering the amount of resources that AI uses (some call it ‘planet-guzzling tech'), it is high time that we stop using it to finish our book reports, create goofy pictures of fictional characters, and entertain us when we are bored.

Responsible use of technology will free up the space for data centres to work on things that benefit humankind while also reducing their carbon footprints.

Old wine in a new bottle?

The world has seen the environmental cost of technology many times before. From the invention of the steam engine to commercial aviation, it is now common knowledge that any advancement increases environmental distress.

This was also on display during the era of cryptocurrency. Crypto-mining also consumes massive amounts of energy, often sourced from resources like coal.

Illegal ‘crypto-farms’ in countries like China became massive sinks of precious resources.

This was strongly opposed and caused governments to take action against such establishments.

But have we learned nothing from history?

Is there a way for us to climb out of this AI-generated hole we find ourselves in?

Way Forward

Apart from the sustainable individual use of AI, it is also essential that big tech companies be forced to reveal their carbon footprintbefore they launch shiny new products.

But “it would be fanciful to suggest that companies invested in developing artificial intelligence with the overarching goal of maximising profit by trying to cut down on payroll are at all motivated by the climate emergency” says Maulik.

Watch:Google blames AI for its carbon emissions | WION Tech It Out

A global effort must be made to treat AI like a precious resource to be savoured rather than gorged, in the form of ‘AI face filters’ and ‘AI food-ordering bots.’

Greater transparency and responsible usage are tenets on which AI can remain a global asset and not another burden on our planet.

We can also go back to asking ChatGPT for the ‘five best breads in the world’ while hundreds of thousands go thirsty and hungry.

The choice is ours.