The environmental impact of AI is multifaceted and depends on how the technology is harnessed and integrated into various sectors. Efforts are being made to develop and promote AI applications that prioritize sustainability, energy efficiency, and positive environmental outcomes. It’s crucial to consider the broader implications and ethical considerations in the development and deployment of AI technologies to ensure a balance between technological advancement and environmental stewardship.
The ‘creative ambiguity’ of the above paragraph, with which ChatGPT sums up its response to a question about the environmental impact of AI development, may in part be a result of its ‘intellectual’ limitations, but it is also indicative of the kind of information with which its authors feed it. That is to say, while they have created a technology with immeasurable environmental impacts, without being able to offer any concrete solutions to mitigate them, they are trying to convince us that the technology itself can ultimately solve the problems it creates.
According to some estimates, the AI industry could consume as much energy as a country the size of the Netherlands by 2027. At the same time, however, the creators of the new technology claim that, as it improves, it can play a crucial role in the design of environmental protection policies (from research into the design of energy-sustainable cities, to studies into the more efficient and less polluting use of renewable energy sources, to the monitoring and protection of sensitive ecosystems, endangered species, etc.).
But the problem facing the planet today has nothing to do with the lack of scientific knowledge on environmental issues. It is anchored to the complete indifference of governments and big business in this area. So even if new AI-type technologies can help to improve environmentally friendly policies, the question is who will implement those policies.
Huge energy consumption
Sam Altman, CEO of OpenAI, has said that a major breakthrough in energy production is needed for AI technology to advance, and has pinned hopes for its future on advances in nuclear fusion. In practice, AI is becoming more and more energy-intensive, while the new methods of producing the energy needed for such leaps are moving at a much slower pace. So AI is using oil, coal, natural gas and some renewables, putting even more pressure on the climate.
The main reason why AI consumes large amounts of energy is that ‘training’ a new model requires the constant processing of large amounts of data, which can take anywhere from a few minutes to several months, depending on the complexity of the model in question. Energy is also consumed by the operation of the AI models themselves. For example, it is estimated that the production of a single image by an AI machine consumes as much energy as it takes to fully charge a smartphone.
Alex de Vries, of Vrije University in Amsterdam, describes that if Google decided to turn its search engine into a tool similar to ChatGPT (so that 9 billion conventional searches per day were turned into 9 billion interactions with an AI engine), the company would end up using as much energy as Ireland just to run this new search engine. In practice, Google would use 29.3 terawatt hours per year for this operation, while its total energy consumption was 18.3 terawatt hours in 2021.
Of course, de Vries explains that the above scenario is not possible in today’s conditions, as it would require an investment of $100 billion to upgrade the company’s machinery to make this transition possible, as well as the ability to procure that volume of machinery, which does not exist today. It is, however, an example of the enormous energy needs of AI.
Water, mining and waste
At the same time, it is estimated that a simple conversation between a user and an AI model like ChatGPT3 uses about half a litre of water, and that Google’s Lamda model uses a million litres of water in its ‘training’ phase alone. Microsoft has admitted that in 2021 and 2022, the years of its big shift to AI, its water consumption increased by 34%. In 2022, it used 6.4 million cubic metres of water, mainly for the cooling of its data centres. This is the equivalent of 2,500 Olympic swimming pools.
At the same time, the continuous development of AI models requires the construction of ever more powerful computers and microchips. This requires more and more mining of the metals used in their manufacture, while at the same time old machines are being replaced faster and faster, piling up in mountains of e-waste, which, according to the World Health Organisation, is the fastest growing solid waste stream in the world. In 2019 alone, a total of 53.6 million tones of e-waste was generated. Of this huge volume, only 17.4% is known to have been collected and recycled. Among other things, e-waste that is not recycled releases toxic chemicals, such as heavy metals, into soil and water.
Will AI help the environment?
Today, representatives of major technology companies that have invested huge amounts of money and expectations in the development of AI promise that the improvement of the models they have built can make a significant contribution in the future to a number of areas related to environmental protection.
According to its creators, the development of AI could help research into renewable energy sources, contribute to the understanding of weather phenomena and their prediction, and therefore indicate the most appropriate forms of renewables for each region. It could help ensure the correct use of this energy and minimise waste in the design of housing, industrial plants and agricultural production, so as to use as few natural resources as possible. In agriculture and livestock production, AI models could analyse soil and climate conditions to improve production and reduce environmental impact. Robots and drones could even reach areas inaccessible to humans, helping to map, understand and protect ecosystems and endangered species.
Even if the above is true or could be true in the future, it could only actually work in a society that does not put profits and competition first, but rather the protection of life and its quality for all humanity, as well as the protection of the environment. A society that will plan its own food and energy production, transport and the manufacture of technological products, based on its own needs and the understanding that the planet’s resources are not inexhaustible.
Such a society would not build a new technology with huge energy needs without first ensuring that these needs would be met from renewable sources and with the least possible impact on the environment. On the contrary, the absurdity of capitalism today has built a “black hole” in terms of energy and water consumption, contribution to emissions, production of toxic waste, etc, and hopes that one day the problem will be solved by the technology itself, which will become smart enough to clean up the mess that its creators left behind.
For now, however, AI remains a tool that – despite theories of it acquiring a mind of its own and overriding its creators – basically does what it is asked to do. And what it is currently being asked to do by the big tech companies is to maximise their profits, even if that means uncontrolled growth with incalculable environmental consequences. This situation will not change as long as not only AI itself, but the entire industry and economy is in the hands of a handful of speculators.
