The Bigger Picture of Climate Change and AI

In the face of escalating climate change, the integration of artificial intelligence (AI) technologies emerges as a compelling solution, despite the ethical and environmental implications it raises. But stringent regulations and a more conscious utilization of the technology are imperative to ensure that the potential benefits of AI in combating climate change are harnessed responsibly and sustainably.

—

In the latest WEF Global Risks report, climate change and risks related to artificial intelligence (AI) topped the chart for the top 10 global risks in the coming decade. The report also points to the interconnections of economic, geopolitical, societal risks with environmental and technological risks. 

As the world navigates short-term shocks from inflation and wars, it is more critical than ever that we do not lose sight of the long-term risks that could upend our future.

The Hidden costs of AI: Environment and Ethics

AI has existed since the 1950s, but with the launch of ChatGPT in November 2022, it surged into the mainstream. Within two months, the chatbot attracted more than 100 million active users. But while ChatGPT has spotlighted AI’s potential to boost efficiency, increase access to knowledge, and reshape entire industries, the technology has detrimental impacts on the environment and raises ethical concerns associated with its training and deployment. 

OpenAI CEO Sam Altman recently revealed that just saying “please” and “thank you” to ChatGPT adds tens of millions in computing costs due to higher energy use. 

The environmental impacts of AI stem from energy consumption in training the AI models, inference from daily use of AI tools, water usage to cool the data centres that power it, and hardware carbon footprint. OpenAI CEO Sam Altman recently revealed that just saying “please” and “thank you” to ChatGPT adds tens of millions in computing costs due to higher energy use. 

Open AI reportedly consumed some 1,287 MWh of electricity to train its GPT-3 model – the equivalent to the energy needed to power over 120 US homes for a year. Due to the sheer volume of queries processed daily, inference accounts for over 60% of AI’s total carbon footprint. 

A study on the water footprint of AI highlighted that depending on when and where AI is deployed, GPT-3 consumes a 500ml bottle of water for roughly 10-50 medium-length response. The same study also found that the water withdrawal from global usage of AI is projected to reach 4.2-6.6 billion cubic meters in 2027, exceeding the total annual water withdrawal from Denmark by 4-6 times.  

Despite these impacts, there is still no standardized method to measure AI-related emissions due to the lack of transparency from providers, variability in the carbon intensity of local power grids, and the diversity of AI tools in use.  

More about the topic: The Green Dilemma: Can AI Fulfil Its Potential Without Harming the Environment?

As AI becomes more embedded in business processes, other ethical concerns are surfacing. These include data privacy, lack of transparency and accountability for AI-driven decisions as well as job displacement. Despite debates on ethical concerns of large language models (LLMs) that are trained using copyright materials, relatively few organisations and individuals have chosen to boycott these tools, suggesting that business and public priorities may favour rapid innovation, productivity, and convenience over upholding intellectual property rights, when these are not seen as directly affecting them.

In a 2024 MIT Sloan Management Review Survey, 52% of respondents said they believe that organizations are not sufficiently expanding their risk management capabilities to address AI-related risks. Transparency and accountability issues also arise when AI technologies are known to operate as “black boxes”, leaving no room for human interventions when it generates outputs without clear explanations of their reasoning. Data manipulation is another challenge, where AI models are trained using biased or manipulated datasets and will therefore produce biased or manipulated results. 

The dilemma over whether AI will replace all jobs reflects a tension between technology disruption and opportunity. According to a 2023 Goldman Sachs report, up to 300 million jobs globally could be affected by the rise of AI, raising questions about business leaders’ responsibilities in workforce transitions. However, the report further highlighted that “although the impact of AI on the labor market is likely to be significant, most jobs and industries are only partially exposed to automation and thus more likely to be complemented rather than substituted by AI.” 

A more recent MIT study echoed a similar sentiment, suggesting that human expertise is still crucial in areas involving small or biased datasets, unexpected situations, or moral decisions.

Facing Climate Change With Realism

At the same time, climate change adaptation and mitigation measures are not moving as fast as needed, mostly due to geopolitical distractions and inconsistent policy commitments. 

When it comes to climate change, the debate often feels binary. On the one hand, scientists are calling for a complete ban on fossil fuels – the main contributor to global emissions. On the other hand, skeptics argue that a total decarbonization is not possible due to costs, lack of infrastructure and its potential impact on the job market. 

Wind turbines at the Shiloh IV wind energy project in Solano County, California. Photo: Wikimedia Commons.

The largest emitters of greenhouse gases, which include China, the US, India, the European Union, and Brazil, have started to incorporate renewables in their energy mix. However, research shows that they may not be sufficient to avoid the 1.5C global warming scenario, which requires at least 80% clean energy mix by 2030 and 100% by 2050. Only a few countries, including Iceland, Norway, New Zealand, and Denmark, are close to achieving this. 

But while the climate transition comes with trade-offs – economic restructuring, job disruptions and high upfront costs – the cost of inaction is far greater, and will be felt disproportionately in poorer and more climate-vulnerable regions that contribute the least to global greenhouse gas emissions.

An International Monetary Fund study estimates that decarbonization can yield a net gain of up to US$85 trillion. Stanford researchers show that moving to 100% clean energy could create a net increase of 24.3 million new jobs, far more than the losses expected in the fossil fuel sectors. We are already seeing this shift, with rising demand for sustainability professionals and training being integrated into education systems. 

The climate crisis requires realism in recognizing the practical steps needed to transition responsibly. That means investing in clean energy infrastructure, supporting upskilling and retraining programs, implementing fair carbon pricing, and fostering international cooperation to mobilize climate finance for the Global South. 

The Bigger Picture 

The evolution of global climate efforts – from the Paris Agreement to the UN’s Sustainable Development Goals – shows that collective actions by governments, business, and international coalitions can facilitate meaningful progress. The European Union and South Korea’s recent adoption of the AI Act and the AI Basic Act, respectively, marked a turning point in global efforts to govern emerging technologies. The new regulations lay down comprehensive rules governing the development, marketing and use of AI in the respective jurisdictions.

In the pursuit of smarter AI models and urgent decarbonization goals, two key questions arise: What if AI succeeds in replacing key human decisions? What if we fail to curb climate change in time? These are not just hypothetical scenarios, but plausible impacts that are already unfolding in parts of the world today. 

To navigate this dual transition responsibly, we need to move beyond binary thinking – between idealism and denial – and embrace a more pragmatic, risk-aware mindset. 

For individuals, this means being more mindful of the frequency and necessity of your AI use, opting for lighter models or carbon-conscious platforms where possible. For developers, it means prioritizing model efficiency, greening data centre infrastructure, and providing transparent carbon reporting. For policymakers, it means addressing the AI and climate risks more holistically, from data privacy and IP rights to energy use, workforce transitions, and governance.

About the Author