Artificial Intelligence, it is an up and coming phenomenon, by now everybody has heard of it. The biggest of course being ChatGPT. Most people I know use it, even if it is out of sheer curiosity. But in the last year we see more and more AI’s popping up everywhere. Samsung has developed an AI, called Gemini. It is the only thing you see, when you watch an advertisement of Samsung. Even when you put in a simple google search, you get an AI that answers your question, before you have to scroll down to look at the results of your search.
As I said, AI is becoming more common in our ordinary life. Is that something we should want? AI makes our life easier, but it is also a huge burden on our planet, which people may not know. It feels very simple: you ask a question and you get an answer. But the costs behind the scenes are huge: it is expected that there will be a 70% increase in data-centre energy demand in Western Europe, because of the increase in use of AI.
If you think about it, it is completely logical that an AI-model uses energy, but what kind of energy. Well, firstly it uses a lot of electricity. A lot more than a ‘normal’ search engine, like a Google search. And to make electricity we need fuel, which of course impacts our environment. If we take a look at ChatGPT, the most renowned AI-model. Take for example GPT-3, which was released in July 2020. It was a newer version of ChatGPT which OpenAI developed. It is estimated that it took 1287 megawatt hours of electricity. To put that into perspective: an average Dutch household uses 2500 kilowatt hours a year, which means the development of GPT-3, costs as much electricity as 514 Dutch households use in a year!
Secondly, because AI needs so much electricity, we need a lot of fuel, which leads to a lot of emissions. The development of GPT-3 generated more than 550 tons of carbon dioxide equivalents. That is the same as 123 gasoline-powered passenger vehicles being driven for a year.
Both these number, the electricity and the carbon dioxide, sound like a lot and they are. These numbers sound even crazier when you consider that it took so much, only to build it. Consumer usage is not even considered in that number! So the use of GPT-3 will have led to a lot more emissions.
Thirdly, AI needs a lot of water. It could sound weird to some people, that something online, needs water, but it does! AI uses a lot of data, which are stored in data centres, and these centres need to be cooled. They get cooled with water, a lot of water. If we once again look at GPT-3 and the development of it, we find that a two-week training of GPT-3, to make it ready for release, took about 700.000 litres of fresh(!) water. A large data-centre could use up to 1.8 billion gallons a day. That is 6.8 billion litres a day, approximately the same as a town of 10.000 to 50.000 people uses in a whole year. This water could have better usage then simply cooling off data-centres, but it does not.
Of course AI can also be of great usage for our climate, by helping us in working out how we can become more sustainable. AI could help by finding more fuel-efficient routes on gps-systems, or making sure that planes fly the most eco-friendly route. AI can help predict climate-related events such as floods. In addition to these specific points, AI can also help us save the climate, because it makes it easier to find information about the climate.
By now you know how much the use of AI impacts our environment. So maybe it is a good idea if we all started to use a little less AI. Maybe we should just stick to a simple google search for now, until we have found a sustainable way of using Artificial Intelligence. Or has AI already permeated our daily life to such an extent that we should, perhaps, actually invest heavily in making it sustainable?
https://www.nibud.nl/onderwerpen/uitgaven/kosten-energie-water/#Elektriciteitsverbruik
https://news.ucr.edu/articles/2023/04/28/ai-programs-consume-large-volumes-scarce-water
https://www.eesi.org/articles/view/data-centers-and-water-consumption
https://blog.google/outreach-initiatives/sustainability/report-ai-sustainability-google-cop28/
Recent Comments