When we say AI, you probably say ChatGPT, DALL-E, or Grammarly; tools a lot of people use for school, work, or just recreational use. When we ask the same question to companies, they’ll likely name the AI programs they use for chatbots, personalised ads, and inventory management; all ways to increase sales, decrease work costs, and save time. But nowadays, companies are giving AI a purpose in a way that doesn’t fit those goals: tackling sustainability issues.
Examples
A clear example of companies using AI for environmental reasons can be found in the (fast) fashion industry. Companies like Zara and H&M use AI tiny RFID microchips in the price tags of their clothing to optimise their inventory management, resulting in less waste as they more precisely know how much and fast something gets sold, making it easier to supply more goods without oversupplying. On top of that, they use an AI system, provided by Refibred, to optimise the recycling process by 70% by identifying materials more precisely, making the textiles end up where they should be making recycling more efficient.
A negative impact
Whilst AI can help companies tremendously with sustainable innovation, AI is actually not as environmentally friendly as it might seem, raising the question if this use is not just an attempt to look more sustainable without really being it or if it’s just plain ignorance.
A study by the University of Massachusetts has shown that the training of just one singular AI model has about the same amount of emissions as 62.6 petrol cars use a year.
According to a Dutch data scientist, Alex de Vries, AI will use up about 1.5 million servers in the future, consuming more than 85 terawatt-hours of electricity annually – which is actually more than some entire countries use. The electricity for these servers can’t be won solely by solar panels and windmills (yet), plus, Patrick de Lima from Global-e Datacenter tells us that most data centers have at least two power connections, including a diesel backup generator, and these physical servers use so much heat it has to be cooled 24/7, which in turn also requires a lot of energy. And then, of course, the housing of these servers that potentially need new and bigger buildings. It is difficult to estimate how much space is needed, though, says Lima, as servers are also hosted online and it depends on the host.
Time for regulation
This problem also concerns the EU. Parliamentarian, Kim van Sparrentak, is a big advocate for more regulation of AI in terms of the environment on top of the existing regulation that mostly concerns privacy. Whilst she acknowledges all the good things it can do, and even how important it is, she also points out that the use of AI in these terms is obsolete when it costs more energy than it saves: “If that costs more energy in the end, it is useless and that’s why we need minimum standards for that amount of energy they use, and we need to make sure that these developers, you know, while they’re creating new apps, that they just make sure that they’re programmed in such a way that they use as little energy as possible.”
We live in a world where AI is growing rapidly and is not planning to stop anytime soon. It’s a tremendous help, but it brings its own problems too, so in the case of sustainability it’s not a permanent solution without enough green energy as it impacts the climate.
H&M Group has decided not to answer when we asked their point of view in this dilemma.
