You might have seen some of these titles:
“Creating an AI can be five times worse for the planet than a car”[i]
“Alarming new research suggests that failure to source renewable energy could make data centres one of the biggest polluters in just seven years”[ii]
“Climate change: Is your Netflix habit bad for the environment?[iii]
Do you still wonder what this is all about? Are you scared and wondering what on earth you—or other companies—could do? Ultimately it all boils down to one topic: the use of energy behind digital technology and related services. To create this energy, we often burn fossil fuels like oil, gas, or coal, all of which release greenhouse gases to the atmosphere. For that reason, the most energy-intensive, and the least energy-efficient digital technology is the biggest problem.
So, what are those energy-hungry technologies? And what can be done to solve the problem?
Energy-use—tied to the use of data
To evaluate the environmental impact, and the amount of greenhouse gas emissions caused by digital technology, you need to ask yourself these questions:
- What would the environmental impact be without digital technology X?
- How much energy does digital technology X require: is it energy-intensive? Is it energy-efficient?
- Source of the energy used by digital technology X: coal, oil, gas, nuclear, or renewables
Energy consumption is fundamentally tied to the use of data. For example, when it comes to website design, the more data-intensive a website is, the more energy it consumes per page view. On a website, video streaming requires the most data and energy. Next comes audio files, images, and GIFs. Other examples of useless use of energy and digital waste include downloading documents to the same device multiple times, storing almost identical photos in the cloud, autoplaying videos, and playing music videos when you only need to access the audio. For example, the common practice of playing a music video on YouTube only to access is audio is more energy-intensive than just playing the audio.
However, if you think beyond the individual level and look at the big picture to assess the direct use of energy by digital technologies themselves, most of the energy is used by data centers.[iv]
Because of this, data centers have a vital role in the mission to reduce greenhouse gas emissions. The Condorcet data center in Paris, which is the most energy-efficient data center in Europe, sends its waste heat directly into a neighboring Climate Change Arboretum, where scientists study the impacts high temperatures have on vegetation.[v] AI and machine learning can also help in reducing the energy intensity of data centers: DeepMind AI has reduced the Google Data Centre cooling bill by 40 percent.[vi] When applied on a larger scale, similar results in any energy-intensive environment would have a phenomenal impact globally.
The second biggest contributor to energy-use after data centers is shifting data through the data transmission network. Video streaming services such as Netflix, YouTube, or BBC iPlayer require less energy in the data center, but the energy-use of data transmission networks and the end-user’s devices are much more intensive. To mitigate the energy-intensity, focusing on mobile-first web design helps you avoid loading large data designed for desktop machines, which improves your site’s speed and energy-efficiency. As mobile devices are used more and more often, mobile-friendly website design also makes sense from both the commercial and user-centric perspectives.
Good news: user-friendly interface and website design is good for the planet
Besides focusing on gaining in energy-efficiency faster than energy-intensity, technology companies should be thinking about how to design the services as sustainably as possible. Researchers believe that this will go hand-in-hand with improved user-experience since reducing elements of a service that people are not enjoying or not using essentially means you’re removing something that wasn’t worthwhile to the user.
The graph visualizes how companies could significantly reduce their carbon footprint by investing in the user-interface design!
By opting for this approach, interface designers could take this metric into account and that way, the “digital waste” could be minimized already in the design process, making end-users’ behavior environmentally-friendlier by default. When the user is only interested in the audio, why continue streaming a resource-intensive video?
Conclusion: combine digital technology and AI with “design for environment” approach
Several studies have calculated the collective CO2 emissions of digital technology to be about 1.4- 2% of total yearly global emissions, which is as much as generated by the aviation industry. Despite this worrying comparison, you shouldn’t worry about your habit of watching cute cat videos on the internet or playing music videos on repeat is a significant contributor to the climate crisis. On a research and development, and business level, however, spotting energy-intensive yet useless features on apps, websites and devices, and designing to avoid them, could have a noticeable impact.
Besides, even if data AI and machine learning require a lot of data and energy, combined with the “design for environment” approach, they have the potential to reduce energy-intensity of digital technology and to improve the energy-efficiency of data centers. Optimizing for energy-efficiency will be essential when going forward, and AI might be the right tool for that.
Finally, technology companies could develop an algorithm that would quickly assess any interface design as well as the code underlying it and determine its CO2 footprint. This way “digital waste” could already be minimized in the design process, making users’ behavior environmentally-friendlier by default.
People make unconscious decisions all the time. That is why companies should focus on creating an environmentally friendly way of life a convenient, default option for people. Companies should be responsible for their environmental impact—online and offline.
[iv] Andrae, Anders. (2017). Total Consumer Power Consumption Forecast.