The Climate Consequences of ChatGPT
It is understandable that people flock to a new creative technology, such as ChatGPT, when it emerges – especially when it is free-to-use and an enormous amount of marketing and PR has gone into creating vast media hype.
While there is an ongoing ethical debate about the application of these AI technologies, they also have enormous implications for the Climate Crisis that are all-too-often overlooked.
Today we give you all the information you need to better understand whether playing with AI is worth the climate consequences.
Creating New Emissions
The IPCC is clear in its AR6 Synthesis Report: We must dramatically reduce greenhouse gas emissions and we must do this urgently if we are to maintain a liveable planet for all.
Yet we have not only seen little international effort to curb existing emissions – at the time of publication, greenhouse gas emissions currently stand at an all-time high – but the creation of brand new streams of emissions over the last few years.
Crytocurrencies and NFTs are a great example of that. They are responsible for producing vast new emissions that simply did not exist previously, and consume resources for little-to-no societal benefit. They are the manifestation of late-stage capitalism, and large cryptocurrencies such as Bitcoin have carbon footprints larger than whole European nations.
Cornell University researchers estimated that training GPT-3 consumed 1,287 megawatt hours of electricity, but now that ChatGPT is processing more than 1.8 billion visits per month, its energy consumption and demand for GPUs is ever-increasing.
ChatGPT raced to 1 million users in just 5 days owing to the publicity behind it. As of April 2023, it attracts more than 1.8 billion monthly users. Credit: Statista
Microsoft has revealed that its AI training consumed over 10,000 graphics cards and 285,000 processor cores.
Stanford University PhD research Peter Henderson stated: “We’re just scaling without any regard to the environmental impacts, we can get ourselves into a situation where we are doing more harm than good with machine learning models. We really want to mitigate that as much as possible and bring net social good.”
The honest answer is that there is no public transparency about how much energy AI chatbots consume, or how much more energy-intensive AI image creation tools are.
However, this issue is particularly pertinent in Ireland, where data centres are anticipated to account for 30% of energy consumption by 2030 and energy consumption in data centers skyrocketed by 400% in the final quarter of 2022 compared to the same period in 2015.
This is delaying the country’s transition to renewable energy, worsening air pollution and jeopardising Ireland’s emissions targets – the ones we are already failing to meet.
Training GPT-3 - the technology that underpins the ChatGPT tool - eclipses the emissions of internal combustion engine cars across their whole lifetime. Credit: The AI Index 2023 Annual Report,” AI Index Steering Committee, Institute for Human-Centered AI, Stanford University, Stanford, CA
Thirsty AI
However, the climate and ecological impact of AI tools doesn’t stop with energy consumption and greenhouse gas emissions.
A typical data center consumes around a gallon of water for every kilowatt-hour expended, and our increased reliance on the cloud drove Google to request more than 2.3 billion gallons of water for its data centers in just three US States alone.
Yet consumption of water is even higher when it comes to AI.
Training alone for GPT-3 consumed 700,000 litres (185,000 gallons) of water for cooling – that’s about the same amount of water needed to fill a nuclear reactor’s cooling tower, according to a joint study by University of California Riverside and the University of Texas, Arlington.
To put that into context, an AI chatbot like ChatGPT will consume the equivalent of a 500ml bottle of water for every 25-50 questions that it processes.
This comes at a time when US and global water stocks are already in a precarious position due to the Climate Crisis – with the IPCC reporting in 2022 that roughly half the world’s population is experiencing severe water scarcity for at least part of the year.
The ‘Making AI Less “Thirsty”’ research paper states: “The growing carbon footprint of AI models, especially large ones such as GPT-3 and GPT-4, has been undergoing public scrutiny. Unfortunately, however, the equally important and enormous water footprint of AI models has remained under the radar.”
“AI models’ water footprint can no longer stay under the radar - water footprint must be addressed as a priority as part of the collective efforts to combat global water challenges”
Individual Responsibility
The onus for change in the face of the Climate Crisis is all too often placed upon individuals. In many instances, this allows the companies behind a product or service to avoid scrutiny for the impact that they are having on worsen the Climate Emergency.
However, in the case of AI chatbots, we can make a big difference by understanding that our eagerness to play with them as a novelty has severe consequences for the Climate Crisis. Cut off the demand and the supply lessens.
Yes, those pushing these technologies on us need to be accountable for their actions – and the AI industry needs to be regulated across the board – but we also need to look at ourselves and ask why we are so quick to adopt these technologies without thinking about the climate impact.
Electing not to play with these tools to create AI-generated text or images will help to cut energy and water consumption, while also reducing emissions.
It is undeniable that AI chatbots and image generation tools are already having a big impact on the Climate Crisis, but we can all play a role in minimising that; just by not engaging with it.
Training GPT-3 had vast consequences for energy and water consumption, as well as creating vast new emissions. GPT-4, its successor, is set to blow this out of the water due to being trained on a bigger data set. Credit: Acquisition.com
What To Read Next
RTÉ Refuses To Address Climate Coverage Failings
RTÉ refuses to be transparent or accountable for its continued failings in reporting the severity, scale and urgency of the Climate Crisis
Syndicated Media & Climate Misinformation
Syndicated articles expose a golden opportunity for the media to replicate greenwashing and misinformation, as recent examples in the Irish press demonstrate
Stop Engaging In Climate Imperialism
We explore what climate imperialism is, why it matters in the face of the Climate Crisis, and why you shouldn’t participate in it