Lately, our social feeds have been buzzing with discussions about the environmental impact of AI. It's a hot topic, and for good reason.

But how energy-consuming is AI technology like ChatGPT really? Let's dive into the environmental costs associated with powering and cooling the data centres that make AI possible.

What is AI?

AI, or artificial intelligence, refers to machine systems that can mimic human-like predictions and decisions. One AI tool which might be familiar to you is ChatGPT, an AI-powered chatbot that leverages human-like interactions to respond to questions, assist with writing emails, essays, code and more.

While hailed as a technological marvel for its capabilities, AI faces criticism due to its significant environmental impact. The complexity of AI systems demands substantial energy, contributing to a growing environmental burden. For instance, a single query on ChatGPT releases 4.5 grams of carbon, which is equivalent to the carbon footprint of 21 Google searches!

How is AI damaging to the environment?

As AI technology continues to advance, its environmental footprint is becoming increasingly significant. A major factor contributing to this is the substantial energy required to power and cool the hardware in data centres. Projections suggest that by 2027, the AI sector could consume as much energy annually as the entire country of the Netherlands. By 2030, AI data centres might account for 4.5% of the world's total energy generation.

Why Does AI Consume So Much Energy?

The high energy consumption in AI primarily stems from three main activities:

1. Data Processing for Model Training: Training AI models involves processing vast amounts of data, which demands continuous operation of specialised hardware. These chips often run 24/7 to handle the data load.
2. Model Complexity: More complex AI models require longer training periods. This extended training time translates to higher energy use.
3. User Requests: The sheer volume of user interactions with AI systems also adds to the energy burden, as each request requires computational power.

Sources of Data Centre Emissions

The environmental cost of data centres isn't limited to energy consumption alone. Emissions are generated from:

• Electricity Usage: Running data centres demands a significant and continuous power supply.
• Cooling Requirements: Keeping servers cool requires large amounts of water, adding to the environmental impact.
• Equipment Lifecycle: Data centre hardware has a limited lifespan, and replacing outdated equipment generates considerable electronic waste.

As we continue to integrate AI into more of our lives, it’s crucial to address these environmental challenges. Sustainable practices and innovations in energy efficiency will be key to mitigating the environmental cost of AI's growth.

How Can AI Help the Environment?

While concerns about AI's environmental impact are valid, it's important not to overlook the many opportunities for AI to benefit the environment. From climate modelling to enhancing energy efficiency in logistics, AI's potential for sustainable applications is vast.

AI is already being harnessed to address the climate crisis. For instance, AI-powered climate modelling allows for faster and more accurate weather predictions compared to traditional methods. This capability is crucial for scientists to understand the changes our planet is undergoing and to forecast weather events with greater precision.

Innovative solutions are also emerging to mitigate the environmental impact of data centres. Take Deep Green, for example. This company has strategically placed its data centre beneath Exmouth Leisure Centre to heat the pool. This clever initiative saves the leisure centre £20,000 annually in energy costs and reduces its carbon footprint by 25.8 tonnes each year.

A Good News Story

There are impressive examples of data centres that achieve up to 99% energy efficiency, such as those in Falun, Sweden. These cutting-edge solutions are not only scalable but also have the potential to make a significant environmental impact. For context, Earth Overshoot Day marks the date when humanity's demand for ecological resources exceeds what the Earth can regenerate in that year. By adopting this energy-efficient technology, we could push this critical date back by up to 21 days!

How Does the Impact Compare? 

To better understand the environmental impact of different digital activities, it helps to compare the carbon emissions associated with each to highlight the significant impact of seemingly routine tasks. Here’s a breakdown of how various online and tech activities stack up in terms of CO2 emissions:

• A single Google search generates about 0.5g of CO2e.
• Sending a text message results in approximately 0.8g of CO2e.
• Sending an email typically has a carbon footprint of around 4.5g of CO2e.
• Every 5 minutes of web browsing from a smartphone produces about 5.6g of CO2e.
• Browsing the internet on a laptop for 5 minutes emits around 8.2g of CO2e.
• An hour-long video call on a moderately efficient laptop contributes about 10g of CO2e.
• A one-hour video call on a desktop computer with a screen generates around 50g of CO2e.
• Lastly, one hour of video-on-demand streaming contributes around 55g CO2e to your carbon footprint. 

Should We Be Worried About AI and the Environment?

We’ve always been concerned about the impact of new tech, from streaming services to search engines, and AI is no different. While it’s true that AI uses a lot of energy, there's plenty of hope on the horizon. 

As technology evolves, we can expect to see more efficient AI systems that cut down on energy use and reduce costs. This is exciting news, as it means we’ll likely see greener, more cost-effective AI solutions coming our way soon!

• • •