The impact of artificial intelligence on the environment | Experts’ Opinions

The rise of artificial intelligence (AI) has generated many concerns about jobs being replaced by this technology as well as infringements of human rights. However, not so much is mentioned about its significant impact on the environment. Did you know that a single search using ChatGPT uses 10 times more energy than hitting a search button on Google? This is because it constructs answers from scratch rather than finding them in an index, and this requires much more energy consumption. More specifically, to train ChatGPT, centers where this is undertaken use as much energy as an American household would use over 700 years. In addition, data centers generate carbon dioxide emissions and consume tons of water to cool systems down. To learn more about the impact of the increasing use of AI on the environment, check out the thoughts of some experts below.

DevelopmentAid: How is the increasing use of AI affecting the environment, particularly with regard to energy consumption and carbon emissions?

“The increasing use of AI, particularly Generative AI (GenAI), has significant environmental implications, particularly in terms of energy, water, and carbon emissions. GenAI systems require substantial energy to operate, including cooling processors and electricity generation which increases demand for natural resources. For instance, AI systems consume large amounts of fresh water for cooling purposes, with projections indicating that by 2027, AI-related water consumption could approach that of several European countries. High energy consumption is also required to train large language models, inference, deployment, and data centers. In terms of energy usage, a GenAI-driven search consumes four to five times more energy than a conventional web search. This heightened energy demand contributes to increased carbon emissions. Additionally, the environmental cost of the hardware required for AI involves mining rare earth metals, while the disposal of outdated hardware contributes to e-waste”.

“The increasing use of AI has a direct environmental impact in terms of energy consumption, e-waste, carbon footprint, resource extraction and biodiversity loss. With regard to energy consumption, the environmental impact of AI is strongly correlated with the energy demand for AI training, cooling system efficiency, algorithm efficiency, data center energy consumption analysis, renewable energy integration and hardware optimisation. With reference to carbon emissions, the environmental impact of AI is mainly due to mining activities for the extraction of vital raw materials and, in particular, the rare earths, needed to build AI systems. In fact, mining activities have been found to be the most energy-intensive activities releasing the largest quantities of greenhouse gases into the atmosphere.”

“As businesses and individuals become more and more enthused over AI models like ChatGPT, we cannot afford to overlook the shortcomings of Large Language Models (LLMs). With gaining momentum worldwide for climate change efforts and sustainable development, AI practices like ChatGPT cannot become an exception to the rule of accountability for sustainability. Training, cloud computing and the inference processes of AI based LLMs guzzle up enormous amounts of energy and water at their data centers. Even though the CO2 emissions can be estimated, it is difficult to accurately quantify the energy (traditional vis-à-vis renewal) usage at data centers due to the lack of full transparency. Data centers have been rightfully termed the ‘black boxes’ of industry as measuring their huge carbon footprint still remains a challenge. In addition, increasing AI usage can undo sustainable development efforts by lowering the costs of production for fossil fuels and other such carbon intensive industries.”

“The increasing use of AI, particularly Large Language Models like ChatGPT, has significant environmental implications. The training and operation of these models require substantial energy, leading to high carbon emissions. For instance, training a model like ChatGPT can produce carbon dioxide emissions equivalent to those of nearly five average American cars over their lifetimes. This energy consumption is akin to that of an American household for over 700 years. As AI becomes more integrated into various sectors, the demand for computational resources will continue to rise, exacerbating these environmental concerns.”

“AI is rapidly transforming our world, but this progress is significantly impacting the environment. The training and running of large AI models require massive amounts of computational power, which leads to significant energy consumption. For example, major tech companies like Google have increased their emissions by nearly 50% compared to 2019, largely attributed to the energy demands of AI (according to CNBC). Data centers, which form the backbone of AI, consume huge amounts of electricity, much of it from fossil fuels. While AI’s current energy use is a small fraction of the tech sector’s power consumption – responsible for an estimated 2-3% of global emissions – the concern lies in the potential surge as AI scales up exponentially (World Economic Forum). This reliance on fossil fuels results in considerable carbon dioxide emissions which contributes to climate change. In addition, manufacturing the hardware for AI systems requires vast amounts of resources, including rare earth minerals. The mining of these materials often has detrimental environmental and social consequences. The rapid pace of technological advancement in AI also leads to frequent hardware upgrades, which generates significant electronic waste, containing hazardous materials that can pollute the environment if not disposed of properly. The AI industry and academia need to urgently work on solutions to address these concerns. Adopting circular economy principles can help to minimize waste and resource depletion by reusing and recycling hardware components, which extends their lifespan and reduces the need for new materials. Finally, promoting responsible AI development practices that prioritize sustainability throughout the AI lifecycle, from design to deployment and disposal, is key.”

DevelopmentAid: What can the international community do to reduce the environmental impact of AI while still benefiting from its advantages?

“To reduce the environmental impact of AI while retaining its benefits, the international community must adopt a comprehensive and collaborative approach that involves the AI industry, researchers, and legislators. It is vital that organizations openly measure and disclose their usage of energy and water resources to foster accountability for environmental impact. Independent audits can play a vital role in ensuring adherence to environmental standards. The AI industry must also prioritize the development of energy-efficient hardware, algorithms, and data centers. Projects like the BLOOM model in France demonstrate that it is feasible to build AI systems with significantly lower carbon footprints which highlights the importance of innovative design. Transitioning to renewable energy sources is another critical step, ensuring that the power needs of AI systems and data centers are met sustainably. Researchers have a significant role to play in minimizing AI’s ecological footprint. Collaborations between AI researchers and environmental scientists can ensure that technical designs align with ecological sustainability goals. Furthermore, the establishment of standards for assessing AI’s environmental impact and the creation of reporting frameworks would encourage developers and operators to adopt sustainable practices. Policy makers and legislators must also step in to regulate and incentivize environmentally conscious AI development. Establishing benchmarks for energy and water usage is essential, along with requiring thorough environmental reporting and impact assessments for AI initiatives. Providing incentives for the adoption of renewable energy and energy-efficient technologies can further encourage sustainable practices across the industry. Through these combined efforts, the international community can address the environmental challenges posed by AI while ensuring its continued benefits for society.”

“The international community should implement concrete mitigation strategies, such as the design of ‘sustainable computing’ through the creation of low-power hardware and optimized algorithms for energy efficiency, the creation of AI systems that can improve sustainable supply chain management processes, improvement in the use of biodegradable materials, and the setting up of green data centers. Policymakers should adopt policies and regulations that promote international cooperation and encourage sustainability and data transparency. For example, adopting effective e-waste regulations and imposing limits on GHG emissions might push the AI industry towards more sustainable practices, just as greater public engagement in educational campaigns, community initiatives, digital literacy and more effective stakeholder engagement can raise the level of community awareness and promote a culture of sustainability.”

“AI has immense potential to boost economic activity and lead us to innovations that are more technical. However, there should be a tempering of the hype surrounding AI systems like ChatGPT and a recognition of the environmental costs of such LLMs. One-way of achieving this would be to better assess the energy-performance trade-offs by benchmarking energy usage. International research should be directed towards ‘greener’ AI models rather than larger and more complex ones. The international community should prioritize greater transparency and climate positive innovations in AI. Individuals and unbiased journalism also have a role to play by recognizing and questioning the environmental impacts of AI practices. Social impacts should also be taken seriously along with the environmental impacts of AI systems. Last but not the least, to achieve sustainability, AI should be controlled by humans and not the other way round.”

“The international community can take several steps to mitigate AI’s environmental impact. Firstly, investing in renewable energy sources to power data centers can significantly reduce carbon emissions. In addition, optimizing algorithms and hardware for energy efficiency can help to lower the energy consumption of AI models. Collaboration between tech companies, governments, and environmental organizations is vital to for the development of sustainable AI practices. Promoting transparency in reporting energy usage and emissions can also encourage accountability and drive innovation towards more eco-friendly AI solutions.”

“One possible solution is to adopt a multifaceted approach to mitigate AI’s environmental impact while still leveraging its benefits. This approach should involve collaboration between governments, research institutions, and tech companies, with a focus on promoting energy-efficient AI through investment in research and development. Additionally, embracing circular economy principles, such as designing for durability, reuse, and the recycling of hardware components, can help to minimize waste and resource depletion. Initiatives to promote the reuse and recycling of AI hardware components are essential, such as take-back programs, incentives for recycling, and developing technologies for efficient e-waste processing. Furthermore, accelerating the transition to renewable energy sources for data centers and AI infrastructure involves investing in technologies such as solar and wind power, which can also significantly reduce the carbon footprint associated with AI operations. Lastly, raising public awareness and educating individuals about the environmental costs of AI can empower people to support sustainable practices and advocate for responsible AI usage. Balancing the benefits of AI with its environmental impact requires innovation, regulation, and a collective commitment to sustainability through both circular economy principles and responsible practices.”

See also: The role of AI in bridging inequality gaps | Experts’ Opinions

Searching on ChatGPT for the best way to find good job opportunities in the international development sector will provide several solutions, such as tailoring CVs and cover letters, applying for jobs that match candidates’ skills, approaching organizations directly, and finding consultancy projects. All these are valid approaches but require a great deal of time and energy. Fortunately, DevelopmentAid’s Individual Professional Membership is a one-stop shop for experts in the field. It provides access to the largest job board, with over 6,500 openings in the aid sector, grants and tenders for individuals, contact information for major organizations and donors, salary trends, and much more.