Making the future of work sustainable

John Kerry, the US Special Envoy for Climate, recently explained that the world was "way off track" in maintaining its pledge to keep the rise in global temperatures to 1.5 degrees Celsius – a figure that’s widely accepted as needed to keep climate change under control.

Kerry explained that keeping this aspiration alive will require reductions of emissions of up to 45% by 2030. However, rather than pursuing that path, society is instead heading toward a rise of between 2.5 and 3 degrees. While sectors like manufacturing continue to dominate in terms of energy consumption, the finance industry can nonetheless play a crucial role in curbing emissions.

The lackluster progress comes despite data from KPMG showing that nearly 80% of companies globally are currently reporting some kind of sustainability metrics, with this rising to practically all companies in countries like the UK. This apparent dichotomy is partly explained by a recent report from the New Climate Institute in collaboration with Carbon Market Watch.

Talk vs action

The report analyses the climate strategies of around 25 global companies, with a particular focus on whether the companies track and disclose their emissions, set emission reduction targets, actively reduce their emissions, and take responsibility for unabated emissions via offsetting or other climate contributions. The report also analyses the transparency of companies across each of the four areas.

Between them, the 25 companies produce the equivalent of 5% of global greenhouse gas emissions, and the report finds that while all of the companies had made pledges to reduce their carbon footprint, those pledges were often ambiguous, and actual concrete commitments were extremely limited.

The report found that all 25 companies committed to some form of zero-emission, net-zero, or carbon-neutral targets. However, it’s notable that only three of these companies, namely Maersk, Vodafone, and Deutsche Telekom, have shown a clear and decisive commitment to deep decarbonization by aiming to reduce over 90% of their full value chain emissions by their respective net-zero and zero-emission target years.

Technological footprint

Technology promises to make meeting these challenges even more difficult. For instance, research from Carnegie Mellon University shows that training a standard natural language processing (NLP) model produces an estimated 626,155 lbs of carbon dioxide emissions, which is roughly 5 times what is produced by a car over its lifetime.

The true environmental impact of datacenters remains uncertain, creating a significant obstacle for the industry, electricity providers, and policymakers to make well-informed decisions on the matter. However, what is irrefutable is the substantial impact the industry has already made and the likelihood of it worsening as the exponential growth of data and digital services persists.

The notion that "data is the new oil" has fueled the drive to accumulate as much data as possible, despite the fact that most organizations only utilize a small fraction of it. In fact, a vast majority of data is mere noise rather than useful information, resulting in not only operational costs but also significant environmental and financial expenses associated with storing excessive amounts of data.

Mike Berners-Lee's book How Bad Are Bananas? The Carbon Footprint of Everything popularized this issue, highlighting that our yearly email usage generates up to 40 kilograms of CO2, equivalent to driving a small petrol car approximately 200 kilometers. As the world becomes increasingly reliant on digital services, it’s crucial to address the environmental impact of datacenters and find ways to minimize their carbon footprint.

Generating emissions

This has become even more problematic with the release and subsequent hype surrounding generative AI models, such as ChatGPT. Estimates suggest that ChatGPT emits 8.4 tons of carbon dioxide per year, more than double the amount generated by an individual, which is around 4 tons annually. However, the precise emissions depend on the power source used to run the datacenters – coal or natural gas-fired plants result in far greater emissions than solar, wind, or hydroelectric power. As such, it's challenging to provide exact figures.

A recent study from the University of California, Riverside, highlights the significant water footprint of AI models such as ChatGPT-3 and 4. During GPT-3's training in Microsoft's data centers, approximately 700,000 liters of fresh water were used – equivalent to the amount required to produce 370 BMW cars or 320 Tesla vehicles.

The intensive training process generates a vast amount of heat, necessitating a staggering quantity of freshwater to regulate temperatures and cool the machinery. Furthermore, the inference process by which ChatGPT responds to queries or produces text also consumes a significant amount of water. For a simple conversation of 20-50 questions, the water used is equivalent to a 500ml bottle, highlighting the substantial total water footprint for inference given the model's billions of users.

While efforts are underway to try and make the sector more sustainable, it runs the risk of invoking Jevon’s Paradox, with any improvements in energy efficiency and sustainability more than offset by the huge increases in usage. What’s more, the analysis by the New Climate Institute urges caution when it comes to the energy efficiency of on-premise IT solutions.

Making change

One body that is aiming to take things in a better direction is the Digital Sustainability Alliance, which was created to tackle the sustainability issues surrounding modern computing directly. The Alliance aims to reduce the carbon cost of data by around 80%.

“The Digital Sustainability Alliance has been formed to open the eyes of policymakers, big tech companies, and industry to the stark reality of the carbon impacts of data consumption and the urgent need for digital sustainability,” says co-founder Ben Golub, the CEO of Storj.

Their aim of reducing the carbon cost by 80% would be sufficient to meet Kerry’s aims, but the industry is currently a long way away from that. Only time will tell if the Alliance has any impact on moving the needle at the speed and distance required to make a difference.