Experimenting with AI: Estimating the Carbon Impact of Digital Solutions

As the world increasingly focuses on sustainability, quantifying the carbon impact of digital solutions has become crucial. Supporting customers in this area, I have recognized a need for more streamlined and comprehensive guidelines and tools. Recently, I came across the European Green Digital Coalition’s methodology published in April 2024 for measuring the climate impact of digital solutions.[Link] This aligns perfectly with the need for a systematic approach to assess environmental impacts of using digital solutions.

However, from my experience with environmental and social life cycle assessments, I know how complex and time-consuming these processes can be. It often involves assembling teams, gathering data, making assumptions, and writing detailed reports—work that can take months and is sometimes only read by a handful of people. Most stakeholders, however, just want a straightforward number or percentage that reflects the impact of their initiatives.

The Burning Question: Could AI Simplify This Process?

As I read through the methodology, I could not help but wonder: How many people will actually read this document, aside from those tasked with quantifying these impacts? It made me think about a more efficient way to approach this. Could AI, specifically generative AI, simplify and expedite the process of estimating the carbon impact of digital solutions?

Curious, I decided to run an experiment using the very famous ChatGPT— known for its ability to streamline tasks and boost productivity. Could it really handle something as complex as measuring the carbon impact of a digital transition? I decided to find out by conducting a hypothetical analysis.

The Experiment: Comparing Luxembourg’s Tax Filing Systems

For this experiment, I used ChatGPT to analyse a hypothetical case - the carbon impact of shifting Luxembourg’s tax filing system from a traditional offline process to a new online platform. The aim was to see if AI could effectively apply the European Green Digital Coalition’s methodology to provide actionable insights. Spoiler: It did!

Here’s how the experiment unfolded:

Step 1: Define the Assessment Scenarios

Scenario A (Traditional Offline Filing): Taxpayers file paper forms, which are processed manually. This involves printing, mailing, physical storage, and possibly travel to tax offices.
Scenario B (Online Filing Platform): Taxpayers file online, reducing paper use and travel but increasing energy consumption due to the online platform’s servers and user devices.

Step 2: Identify and Calculate Effects

Direct Effects for Scenario A: These include emissions from paper production, postal services, and travel.
Direct Effects for Scenario B: These include energy consumption for the online platform and reduced emissions from decreased paper use and travel.

Step 3: Calculate the Carbon Impact

It used hypothetical data to estimate the carbon impact of both scenarios and compared them to determine the net impact. (I provided only two data points: Luxembourg's population of 600K and the goal is to achieve 80% online tax filing with the new platform.)

Results: The Carbon Impact of Going Digital

Based on assumptions about Luxembourg’s population and typical emission factors, here’s what ChatGPT came up with:

Scenario A (Offline Filing): Total emissions were approximately 1,163,400 kg CO₂ annually. This accounts for emissions from paper production, postal services, and travel.
Scenario B (Online Filing): Total emissions were approximately 258,136 kg CO₂ annually, including the energy required to maintain the online platform.
Net Impact: Transitioning 80% of tax filers to the online platform would result in a reduction of 905,264 kg CO₂ annually. WoW !!!

A Curious Thought: Did Using AI Increase My Carbon Footprint?

After seeing these results, I started wondering: Was my carbon footprint higher because I used AI to conduct this experiment? Or would it have been more environmentally friendly to calculate everything manually?

To find out, I compared the energy and carbon footprint of using ChatGPT to perform this analysis versus doing it manually (again using ChatGPT).

AI-Assisted Calculation: ChatGPT consumed about 0.0003 kWh of energy, with a carbon footprint of 0.000069 kg CO₂ (that’s 0.069 grams!).
Manual Calculation: Assuming I used a standard laptop for one hour, it would consume about 0.05 kWh of energy, resulting in a carbon footprint of 0.0115 kg CO₂ (11.5 grams).

The AI-assisted method was much more energy-efficient and environmentally friendly for this specific task. However, the broader lifecycle impacts of AI, including the energy used in training and maintaining the models, should also be considered.

Conclusion: The Role of GenAI in Sustainability

Understanding the environmental impact of digital solutions is crucial for informed decision-making. However, conducting such an analysis can sometimes take months! That is where generative AI (GenAI) comes in. While the data still needs to be checked for quality and accuracy, tools like ChatGPT—or custom-built GenAI chatbots—can significantly reduce the time required to arrive at results, making them invaluable assistants in our sustainability efforts.

As we continue to explore the applications of GenAI, I have found an interesting one for sustainability. If you have come across any GenAI tools specifically designed for environmental impact assessments, I would love to hear about them—please share in the comments!