Self commissioned

2024

Whitepaper over de schadelijke gevolgen voor de omgeving van ons dagelijkse digitale gedrag

SERVICES

Research, writing, illustrations

Agency

Self commissioned

In the modern era, where digital technology permeates every aspect of our lives, we rarely consider the environmental cost of our digital behaviors. From everyday online activities to data storage and transmission, our digital habits have a significant carbon footprint. Data, while often seen as intangible, requires vast amounts of energy to store, process, and transmit. In fact, if the internet were a country, it would rank as the sixth most polluting nation in the world. Also, the estimated global CO2 emissions of data and cloud storage now already exceeds the carbon footprint of aviation. This paper delves into the impact of our digital habits on the environment, particularly the carbon footprint of data usage, the carbon emissions from artificial intelligence (AI), and the responsibility we have as individuals, companies, and creators to minimize these effects. Read the whole white paper online by clicking the title, or read the summarized content below.

Why data is so poluting

Data is not simply stored "in the cloud"; it resides in physical data centers that require a massive amount of energy to maintain. These data centers, which number around 8,000 globally, are powered by electricity, and their carbon emissions vary significantly depending on the country’s energy grid. For example, a kilowatt-hour (kWh) of energy in Germany produces 385 grams of CO2, while the same amount in India results in 632 grams of CO2.

Data centers consume enormous amounts of energy, with energy demand doubling every four years, making the digital sector the fastest-growing energy-consuming industry. In the Netherlands alone, data centers used 13 billion kWh in 2020, releasing 6 million tons of CO2. The growth of digital data usage is projected to contribute 14% of global CO2 emissions by 2040, underscoring the urgent need to rethink how we use and manage digital tools.

The huge carbon footprint of Artificial Intelligence (AI)

All Artificial Intelligence (AI) models, such as GPT-3 and GPT-4 know two phases: a training phase, where the model is trained, and an inference phase, where the model is used by end-users.

In the training phase, the models are fed with huge amounts of data and analyses the data by a vast amount of parameters, a process that demands enormous computational power. In general, the larger the models (the so-called LLM), the more computational power is needed to train the models. Let’s use Chat-GPT as an example: GPT-2 had 1.5 billion parameters, GPT-3 has 175 billion and GPT-4 an estimated 1 trillion (1,000 billion).

For instance, the estimated carbon emissions of training GPT-3 is about 502 tons of CO2, which is equivalent to the emissions generated by 112 gasoline-powered cars for a year. Estimations of the energy use of the newer GPT-4 are around 40 (!) times higher than GPT-3. And AI models need to be retrained several times to keep the outcomes accurate.

You might think the training phase is the most energy consuming phase for AI, but that might not be the case. Google reported that 60% of AI-related energy consumption stemmed from inference.

Artificial Intelligence is booming and companies are lining up to develop their own AI application or to see how they can integrate AI in their own working processes. The adoption of AI in digital tools we use everyday might seriously ramp up our global daily energy consumption... without the majority being aware of this.

So think before you prompt.

Now what?

We live in an age where the destruction of our climate and ecosystems become terrifyingly tangible. It calls us to reflect upon our daily habits, the things we do and the resources we use, either directly or indirectly.

At the individual level, each of us contributes to the digital carbon footprint, whether we realize it or not. Everyday actions like sending emails, watching Netflix, or taking a picture on your mobile (which is saved in the cloud) can add up, producing measurable emissions.

For companies, the digital tools used for collaboration and production contribute to the overall carbon footprint. Many companies use a dozen of digital tools, each with its own database and emissions. Companies can reduce their environmental impact by assessing the carbon footprint of the tools they use and optimizing their workflows, without turning to AI for each optimization.

While AI can offer significant advantages, its environmental costs must be taken seriously and must be taken into decisions about its adoption.

Ultimately, individuals and businesses must take a proactive approach to reducing their digital carbon footprints, embracing more sustainable practices, and innovating for a greener future. As we continue to develop and rely on digital technologies, it is essential to consider their environmental impact and work toward mitigating the damage they cause to our planet.