Global warming and AI: Urgent need to curb AI's environmental impact

Global warming and artificial intelligence (AI) are two of the hottest discussion topics among scientists, technologists, economists, and policymakers today. They are also discussed animatedly in the corner offices of big corporations and the drawing rooms of the cognoscenti.

 

Except for die-hard climate change deniers, almost everyone agrees that global warming is a serious threat — to life and the economy, globally and locally. Global warming and climate change also offer many business opportunities — as all disasters inevitably do. But most people agree that the business opportunities of climate change are nothing compared to the threat it poses to humanity.

 

Meanwhile, recent developments in AI evoke mixed emotions — the potential of it revolutionising industries and supercharging the economies for those who can use it properly is tantalising. The dangers too are apparent — from the job losses that it will inevitably cause to the deepfakes that are destroying lives, influencing elections and generally causing havoc.

 

Global warming and artificial intelligence are rarely discussed together or even considered connected themes. But this is a mistake because they are linked closely. The history of global warming shows that while carbon emissions started when humans discovered how to make fire, it was really the technology revolutions that truly brought us to our current predicament. Each technology revolution — from the steam engine to thermal electricity to the hydrocarbon revolution and the rise of information technology — increased the quantum of emissions exponentially.

 

The balance between natural carbon sinks and carbon emitters started changing. As forests were cleared and water bodies exploited, the capacity of natural carbon sinks diminished. And as we generated more electricity by burning coal and hydrocarbons and developed an insatiable appetite for petrol and diesel in personal transport as well as industries, the global warming clock started ticking faster. The inevitable side effect of development and better living standards and incomes has been more emissions.

 

But all these — hydrocarbons, coal, the internal combustion engine, cement and steel plants, and giant cargo ships consuming tonnes of fuel powered by dirty energy sources — are well-recognised villains.

 

Less discussed and vilified is the role of digital technologies and the information technology revolution — the ubiquity of personal computing and mobile computing devices, the rise of the internet, the Cloud revolution, the surge in the popularity of cryptocurrencies, and now, the Generative AI revolution.

 

The data centres that make our lives easier at home and in the office are huge consumers of electricity and water. They contribute significantly to global warming and emissions. They heat up the environment directly, despite air conditioners used to keep them cool. It is the nature of the computing process that generates heat. Giant data centres emit immense heat, and Big Tech has often looked for solutions such as creating data centres underwater or setting them up in extremely cold geographies. While these methods may keep the servers cool, they ultimately add to global warming.

 

The problem has escalated with the popularity of Generative AI and Large Language Models (LLMs). The energy consumption of this latest technology is gargantuan. A recent study estimates that the daily electricity usage of ChatGPT is equivalent to the daily consumption of 180,000 US households. And that’s just ChatGPT. Its competitors — from Google Gemini to Anthropic’s Claude, Meta’s Llama, X’s Grok, France’s Mistral AI, and others—are likely consuming similar amounts of energy. Then, there are the GenAI companies in China that are trying to outpace the Western world’s GenAI champions.

 

Optimists argue that AI itself will help solve the global warming crisis. Many policymakers also advocate for using renewable energy sources like solar power to run AI cloud centres. But the energy usage of AI is outrunning any renewable energy capacity addition. In desperation, a large number of Big Tech companies — from Microsoft to Amazon and OpenAI — are now turning to nuclear power. From reopening the shuttered Long Mile Island nuclear facility to investing in modular nuclear fission projects, and betting on nuclear fusion startups, the Silicon Valley AI tycoons are exploring all nuclear options.

 

Will these help? It remains to be seen whether nuclear will be a boon or bane. The reason the world fell out of love with nuclear was its dangers — and there is little to suggest that those dangers have vanished.

 

Can we turn the clock back and put a halt to AI developments until a solution to global warming is found? That is silly, as humankind cannot go back in development.

 

While there are no immediate solutions in sight currently beyond what is being tried out, policymakers and technologists need to start the discussion on the AI revolution and accelerated global warming as being interlinked, and mandate that AI companies first figure out how to not add to carbon emissions, even before they propose new solutions for other industries. Without that, it is unclear whether AI will destroy mankind directly by taking over all of society, as some science fiction writers have predicted, or whether it will lead to the extinction of most life on Earth through its contribution to rapid global warming. Neither option is comforting.

 

The author is former editor Business Today and Businessworld and founder of Prosaic View, an editorial consultancy