Skip to main content

Generative AI: Boon or bane for the planet?

Exploring the applications and risks of generative AI for environmental sustainability

Generative AI - or ‘Gen AI’ - the words on every innovator, tech leader and future-savvy person’s lips as we head towards 2024.

It’s been no different in the Deloitte Market Innovation team. As a team dedicated to emerging trends, catching waves and translating innovation into business value, it’s no surprise that it’s hard to go through a team meeting without generative AI featuring in some way, be it marvelling at the tech capabilities, ideating around applications, or discussing the ethics and practicalities of use.

Since a number of generative AI models and products were released this year – including OpenAI’s ChatGPT and DALL·E, Google’s Bard, and many others - it’s a topic that has taken us, and arguably the world, by storm, with opportunities across sectors and professions.

The on-going environmental crisis


We are also in era dubbed the ‘Anthropocene’, characterised by climate change, biodiversity loss, pollution, natural habitat loss, crop failures, and more. Our planet is in crisis.

We have speculated for years that technology will help us face these challenges, and many technologies have and continue to do so.

For example, we can use gene editing to create more resilient crops, develop renewable energy sources to reduce fossil fuel emissions, monitor deforestation from satellites to reduce habitat loss, create better climate models to mitigate and plan for climate impacts, direct carbon capture to remove atmospheric CO2, baseline biodiversity using sound and image recordings, and more.

AI has also already been applied extensively to sustainability tasks, but can generative AI help save our planet as we know it?

Generative AI as a tool for enhancing environmental sustainability


Generative AI surpasses the previous evolutions of AI because it is ‘generative’ – it can come up with new ideas, concepts and solutions.

This opens up many new applications, including for environmental sustainability. For example, we could use generative AI to create future climate scenarios, and then bring these scenarios to life to affect behavioural change. We could also use generative AI to help code better climate models, simulate responses to climate policies, improve manufacturing processes, and improve the interpretability of and access to climate risk analysis to support better decision-making at corporate and governmental levels. The possibilities seem vast, diverse and useful.

To examine this area further, in this blog we’ll delve deeper into three specific ways in which generative AI can be used to enhance environmental sustainability that we as authors think are especially interesting and relevant. We’ll then also explore some of the environmental challenges that are created by this new technology.

Architecture and city planning – towards more sustainable homes, offices and built environments

Generative AI can be used to help design more sustainable buildings and urban spaces.

For example, architect and professor Andrew Kudless has integrated generative AI into his design process, generating over 30,000 images. He says: “If you go into it [generative AI] knowing exactly what you want, you’re going to be disappointed.”Conversely, the benefit of generative AI is that it has the potential to provide the user with something novel, that they haven’t thought of before, and that is not constrained by their preconceptions.

While Kudless is focusing on aesthetic design, this same point carries to sustainable design. An architect could, for example, integrate generative AI into their workflow to help make their designs more energy-efficient and environmentally sustainable. Generative AI could analyse existing designs and suggest modifications, and an architect could then either integrate these changes, or at the very least be inspired by them.

This idea goes beyond individual rooms and buildings to entire towns and cities. For example, Sidewalk Labs have developed a generative design tool called Delve that uses AI to create urban planning scenarios. By combining geographical information with data on regulations, street layouts, weather, building specifications and more, their tool creates a series of possible scenarios for planners to assess and refine.

This type of analysis could enable planners to generate and compare different designs that incorporate sustainability, allowing them to quickly and efficiently assess the impact of different design options early in the planning process and in potentially novel ways.

Drug and materials discovery – towards rapid discovery for greener, more sustainable R&D processes

In the past three years, venture capital firms have invested more than $1.7 billion into developing generative AI technologies, with the largest investments in AI-powered drug discovery and AI software coding tools.

AI has already successfully identified new drugs in the pharmaceutical industry, but by 2025, it is estimated that more than 30% of new drugs and related materials may be systematically discovered using generative AI.

This uses a process known as ‘inverse design’, which defines the required outcomes and properties of a drug, and then searches for and creates drugs that fit these requirements. This significantly speeds up the drug discovery process, which is typically long (6 to 10+ years) and expensive ($2.6 billion is the estimated development cost for each Food and Drug Administration (FDA)-approved drug).

While this has clear benefits medically and financially, there are also significant sustainability benefits, not least by reducing the amount of ‘trial and error’ when searching for drugs. Inverse design could also reduce resource reliance and improve manufacturing processes by helping to find sustainable chemicals and greener alternatives for drug development.

A similar approach can also be used in materials discovery, which is a field that also often requires extensive R&D and trial and error processes. For example, researchers have used generative AI modelling to accelerate the discovery and design of industrial-scale polymer membranes that separate and capture CO₂ more efficiently at its point of emission, which is an approach that is considered to be one of the most effective ways to limit the release of carbon into the environment.

Energy creation and use – towards a world run with a lower carbon footprint and climate impact

AI is currently used to improve the efficiency and effectiveness of existing and new energy systems through, for example, supply and demand forecasting and grid optimisation. Using generative AI to identify and replicate energy-saving patterns could further improve these efficiencies.

While not strictly generative AI, AI techniques have also been used by scientists to control nuclear fusion reactions within a Tokamak reactor. Researchers trained a deep reinforcement learning model to shape more precisely the reactive plasma, which is one of the most challenging aspects of nuclear fusion. Perhaps the next step is for generative AI to suggest further novel configurations and help create a nuclear fusion reactor that is a commercially viable energy source.

A double-edged sword - the environmental cost of generative AI

Generative AI’s negative environmental impact is often overlooked and overshadowed by the potential positive benefits of the technology. However, the environmental impact is big: one estimate suggests that the development and use of a single generative AI model could have a carbon footprint of 284 metrics tons of CO2e, which is equivalent to the lifetime emissions of 5 average American cars, including manufacturing.

This negative environmental impact has started to gain more attention. For example, the Harvard Business Review recently published a series of recommendations for reducing the environmental impact of generative AI, including using existing models, using energy-conserving computational methods, and evaluating the energy sources of cloud providers and data centres. These are all factors that should be incorporated into any new data and technology strategy for businesses that are using generative AI.

There are no silver bullets

While generative AI is already amazing today, it is still nascent. It is a great tool to support our thinking, development and efficiency, but it can’t solve all sustainability challenges, at least not yet.

There are also important ethical and practical considerations, including job displacement, concentration of power, accentuation of existing discrimination and biases, destabilisation of international strategic stability, security and cyber security risks, further proliferation of fake news and the erosion of democracy, concerns over data privacy, and more. And that’s without touching on the hotly debated topic of existential risk.

As with all tools, the tool itself is not the solution: the tool requires appropriate and effective use. Many technological developments to date have had positive benefits to humanity, but also usually negative sustainability impacts (Figure 1). With generative AI, we need to combine an understanding of its potential negative externalities with its judicious use as a tool, and in doing so, we may have a powerful new solution to help save our planet as we know it today.

If you would like to find out more about our work on generative AI and environmental sustainability, please get in touch.

Innovation/Technology Benefit to people Unintended negative impacts on planet
Combustion Engine Faster travel Huge amount of carbon generation
Plastics An amazing material with millions of applications Ocean pollution and biodiversity impacts
Space exploration Knowledge of our universe, space data and communication Space junk
Cloud computiong Secure storage, scalable compute power, instant acess to greater tools Giant data centres with huge carbon footprints
Industrial farming The ability to produce enough food Extensive use of fertilisers and pesticides
Blockchain Distributed immutable data, secure transactions and smart contracts High carbon footprints
Air travel Increased connectivity, tourisim and standard of living High carbon footprints
Innovations in wild-capture fisheries Increased catch at lower effort, improved food quality Environmental impacts on marine ecosystems, stock collapse
Bio based sustainable avaiation fuel Reduce carbon footprint of flights Increased land use to produce bio based fuels with impacts on ecosystems and food security

Figure 1: Many technological developments to date have had positive benefits to humanity, but also usually negative sustainability impacts

Our thinking