The tools, the products and the services based on artificial intelligence they are more widespread and appreciated every day. But they are being carried forward in parallel studies and research dedicated to AI: more precisely to his impact from an environmental point of view.
As is known theartificial intelligence is powered and supported by highly performing devices: capable of assimilating and process very large amounts of data. And this type of device consumes an equally notable amount of electric energy.
The results from this point of view appear to be shocking. Some sources claim that the Google AI they would require the same energy consumption of a small nation. While a generative AI tool in style ChatGPT would end up consuming every year as much electricity as the whole of Ireland consumes.
To get a clearer idea of the picture you can take a closer look at some recently published articles. Above all, an in-depth piece published in the magazine Joule. And an analysis by the researcher Alex de Vries.
How to understand how much electricity AI needs?
Artificial intelligence requires a considerable electricity consumption. And generative artificial intelligence consumes up to ten times more.
In both cases the main source of consumption is i dedicated servers. Devices that risk skyrocketing the carbon footprint of data centers around the world.
An important alert in this sense comes from the aforementioned Alex de Vriesalready author of studies dedicated to the topic: the impact of cryptocurrencies on the environment.
On the one hand, according to de Vries, it is still too early to precisely calculate thepollution directly resulting from the development of AI.
On the other hand, the researcher invites us to become aware regarding theenvironmental impact of artificial intelligence and generative artificial intelligence.
AI consumption data and forecasts
The analysis of consumption data relating to the recent past is an excellent starting point for making predictions about the future. For example already in 2021 l’AI represented approximately the 15% of consumptionof electricity by Google.
Over the last few years the Mountain View Colossus has invested energy and resources in artificial intelligence andgenerative artificial intelligence.
But, unfortunately, the calculation question increased much more slowly than energy requirements. These are not external journalists who declared this: it is Google’s Head of Sustainability Kate Brandt.
Large corporations are engaged in practices and experiments to reduce the carbon footprint of AI: the goal is reduce energy required to train a model of 100volte. E reduce emissions related to the operation by 1,000 times.
At the same time journalistic sources claim that Google uses over 510,000 servers A100 HGX. More specific data speaks of a supply of 100.000 server l’anno by Nvidia.
These servers could consume from 5.7 to 8.9 TWh of electricity every year. If the number of servers provided reaches 1.5 million by 2027consumption could exceed 134 TWh of electricity per year.
To know more: Artificial Intelligence: what it is and what it can do for us