Setting the stage for Sustainable AI and AI for Sustainability
In a world where technology, environmental and social concerns are increasingly intertwined, the emergence of Sustainable Artificial Intelligence (AI) and AI for Sustainability represents a promising solution to address pressing environmental and social challenges. These two closely related fields hold the potential to reshape industries, mitigate climate change, and usher in a more sustainable future in the broadest sense.
In this article, we will:
- explore what Sustainable AI and AI for Sustainability can lead to;
- highlight their differences;
- provide examples of how they can be applied to create a greener, more harmonious and socially responsible world.
Unraveling Sustainable AI
Sustainable AI, most commonly known as ‘Green AI’ or ‘Eco-friendly AI’, what refers to the practice of developing and deploying AI technologies in a manner that minimizes their environmental impact and maximizes their long-term sustainability. It encompasses various strategies and principles aimed at reducing the carbon and ecological footprint of AI systems throughout their lifecycle, from design and manufacturing to operation and disposal.
The ESG perspective on Sustainable AI
This definition of Sustainable AI from an ESG (Environmental, Social, and Governance) perspective can be defined as:
“The practice of developing, deploying, and using artificial intelligence technologies in a way that aligns with responsible and ethical principles to promote positive environmental, social, and governance outcomes.”
Here's a breakdown of what sustainable AI means in each dimension of ESG and why this is important:
Minimizing carbon footprint & the quest for energy-efficient AI models
Environmental (E): Sustainable AI focuses on minimizing the negative environmental impacts associated with AI technologies. This includes reducing the energy consumption of AI systems, using sustainable materials in hardware components, and minimizing electronic waste. Sustainable AI strives to make AI technologies more energy efficient and eco-friendly.
Researchers (Strubell et al., 2019) calculated that training a medium-sized generative AI model using a technique called ‘neural architecture search’ (a smart system that helps design powerful computer brain structures to make machines smarter at tasks like recognizing images or understanding language) has an electricity and energy consumption equivalent to 626,000 tons of CO2 emissions — or the same CO2 emissions as driving five average American cars over their lifetime. These models need to become more energy efficient.
Ethical AI practices in the social context & the challenge of AI-generated images
Social (S): In the social dimension of ESG, Sustainable AI involves ensuring that AI technologies are developed and deployed in ways that respect human rights, promote fairness and inclusivity, protect user privacy, and contribute positively to society. It emphasizes ethical AI practices, avoids discriminatory outcomes, and harnesses AI for social good.
One disturbing issue that requires attention is the need to restrict the use of AI in the generation of certain types of imagery. Since the launch of OpenAI (OPENAI), there has been a significant increase in the creation of realistic images featuring child nudity, as reported by the New York Times. This situation represents one of the most concerning scenarios for machine learning, and highlights the necessity for increased regulation and awareness, particularly around the social aspects of AI.
Ensuring responsible AI governance & transparency and accountability in AI
Governance (G): Sustainable AI in the context of governance is about establishing robust governance structures and accountability mechanisms for AI systems. This includes transparency in AI decision-making processes, adherence to data privacy regulations, and clear policies for AI development and deployment. Sustainable AI ensures that AI is governed responsibly and ethically within organizations.
AI systems can rapidly and extensively generate fake news, with serious consequences such as spreading disinformation, undermining trust, influencing decision-making, damaging reputations, and causing societal unrest. Addressing this issue requires urgent measures, including improved fact-checking, media literacy, and regulation of the use of AI in information dissemination. The question is: how will this be regulated from a governance perspective?
The good news is... we can harness AI for Environmental and Social challenges
Where the above explanation provides a good view of how AI can be misused. AI has a broader scope to be used to help the environmental or social aspects in ESG. Looking at how Sustainable AI can be used to do good, let’s dive into some examples of how Sustainable AI can be used to:
- Energy efficiency (Environmental): Tech companies like Google [Google, 2020] have made significant efforts to improve the energy efficiency of their data centers. They use AI algorithms to optimize cooling systems and reduce energy consumption, resulting in lower carbon emissions.
- Data sustainability (Social and Governance): Institutions adopt responsible data practices by anonymizing customer data to protect privacy while still extracting valuable insights for risk assessment and fraud detection. They comply with data protection regulations such as GDPR.
- Fairness and inclusivity (Social): Recruitment companies employ AI-driven tools that mitigate bias in the hiring process [WHO, 2021]. These tools ensure that candidates from diverse backgrounds are evaluated fairly, contributing to a more inclusive workforce.
- Ethical AI practices (Social and Governance): Healthcare organizations ensure that AI-driven medical diagnosis and treatment recommendations adhere to ethical guidelines, preserving patient rights and well-being while improving healthcare outcomes [WHO, 2021].
- Robust governance (Governance): Large tech firms [IBM] establish AI ethics boards and committees to oversee the responsible development and deployment of AI technologies. These governance structures ensure adherence to ethical standards and regulatory compliance.
The above examples demonstrate how organizations can integrate the critical principles of Sustainable AI into various aspects of their operations, addressing both environmental and social concerns while upholding strong governance practices. In doing so, they contribute to a more responsible and sustainable use of AI in alignment with ESG values.
Exploring AI for Sustainability
AI for Sustainability refers to the use of artificial intelligence (AI) technologies to address environmental, ecological, and social challenges and promote sustainable development. It involves applying AI techniques to collect, analyze, and interpret data related to issues such as climate change, resource management, conservation, and other sustainability concerns.
By leveraging AI algorithms, machine learning, and data analytics, AI for Sustainability aims to find innovative solutions to complex environmental and inclusivity problems and contribute to a more sustainable future.
Currently, many of these AI solutions focus on the environmental challenges and less on the social aspect.
Possible reasons for this could be:
Measurability & data availability
Environmental challenges are often easier to quantify. There's a wealth of structured data available regarding climate, pollution, and biodiversity. AI excels at processing and analyzing such quantitative data. In contrast, social challenges tend to be qualitative in nature, and the available data may be less structured, making them more difficult to model.
Public and political urgency
Environmental challenges, especially climate change, currently hold a high priority on international agendas. This may lead to greater funding and support for technologies, including AI, to address these challenges. Social issues, on the other hand, can sometimes be viewed as more fragmented or local, potentially resulting in less global collaboration or investment in AI-driven solutions.
Clarity of the problem
While environmental issues are complex, they often have a more clearly defined objective, such as reducing CO2 emissions. Social problems, like poverty or inequality, have multiple facets and interconnected causes, making the formulation of a clear AI solution more challenging.
Though these arguments provide some explanation for the observed trend, it's crucial to note that this is a snapshot in time, and the focus of technological solutions like AI may shift as societal priorities and availabilities evolve.
Now, let us have a look at some key environmental areas.
Key areas within AI for Sustainability include:
- Environmental conservation and monitoring
AI technologies are being employed for environmental monitoring, such as tracking deforestation, monitoring wildlife, and analyzing climate patterns. These applications could be valuable in conserving natural resources and biodiversity, thereby promoting environmental sustainability.
- Reducing unemployment
AI-driven platforms can match job seekers with appropriate job opportunities, reducing unemployment rates. By analyzing skills, experience, and labor market trends, these platforms can help individuals find suitable employment, addressing a significant social challenge.
- Smart agriculture and precision farming
AI technologies enable precision agriculture by analyzing data from sensors, satellites, and drones. This data-driven approach optimizes agricultural practices, minimizing water usage, reducing the need for pesticides, and enhancing crop yields. Sustainable agricultural practices promote environmental conservation. [WUR]
- Waste management and recycling
AI-powered systems are employed in waste sorting and recycling facilities. Robotics and machine learning help identify recyclable materials, enhancing the recycling process. Efficient waste management reduces environmental pollution and conserves resources.
- Combatting poverty
AI algorithms analyze socio-economic data to identify patterns and areas of poverty. By understanding the underlying causes, governments and NGOs can develop targeted programs that provide financial assistance, job training, and educational resources to uplift impoverished communities, addressing a fundamental social challenge.
Differences between Sustainable AI and AI for Sustainability
Sustainable AI is now primarily concerned with the responsible development and operation of AI technologies, with a focus on energy efficiency, data sustainability, and transparency. It aims to minimize the environmental impact of AI systems themselves.
AI for Sustainability, on the other hand, concentrates on using AI as a tool to directly address environmental and social challenges. It seeks to solve sustainability issues, such as climate change, resource management and conservation, by leveraging AI technologies.
Thus: Sustainable AI aims to make AI technologies more environmentally and socially friendly. In contrast, AI for Sustainability aims to leverage AI to directly advance environmental and social goals.
The synergy and complementary nature of the two approaches
Despite their distinctions, Sustainable AI and AI for Sustainability are not mutually exclusive; they complement each other. Sustainable AI provides the foundational principles and practices necessary to reduce the environmental footprint of AI systems. These sustainable AI technologies can then be harnessed in AI for Sustainability initiatives, enabling more efficient and data-driven solutions to challenges.
Navigating the challenges: A path toward a Sustainable future
While Sustainable AI and AI for Sustainability offer promising avenues for addressing challenges, they are not without their challenges. Some of these challenges include:
- Human rights, fairness and inclusivity: Developing and deploying AI in ways that respect human rights, promote fairness and inclusivity, protect user privacy, and contribute positively to society.
- Data privacy and security: Balancing the need for data in AI for Sustainability with privacy concerns is a complex task. Collecting and analyzing environmental data while protecting individuals' privacy requires careful consideration.
- Energy efficiency trade-offs: Striking the right balance between energy efficiency and model performance in AI models can be challenging. Achieving energy efficiency may involve trade-offs with accuracy, requiring continuous optimization efforts.
- Policy and regulation: Developing comprehensive regulations and standards for Sustainable AI and AI for Sustainability is essential to ensure responsible and ethical practices across industries.
- Cost and accessibility: Implementing AI for Sustainability solutions can be costly, limiting access for smaller organizations and developing countries. Efforts are needed to make these technologies more accessible and affordable.
Embracing Sustainable AI and AI for Sustainability
Sustainable AI and AI for Sustainability represent two interrelated approaches to addressing environmental challenges in our technology-driven world. However, there lies a massive opportunity in the social aspect of ESG.
As we move forward, integrating these principles into our technological advancements and global strategies becomes imperative. The synergy between Sustainable AI and AI for Sustainability holds the key to a more sustainable and harmonious world. Embracing these concepts and fostering collaboration will enable us to create a future where AI not only drives innovation, but also safeguards the environmental balance and our quality of life.
Google. (2020, Feb 27). Data centers are more energy efficient than ever. Retrieved from https://blog.google/outreach-initiatives/sustainability/data-centers-energy-efficient/
WHO. (2021, June 28). Ethics and governance of artificial intelligence for health Retrieved from https://blog.google/outreach-initiatives/sustainability/data-centers-energy-effici
Strubell (2019, June 5). Energy and Policy Considerations for Deep Learning in NLP Retrieved from https://arxiv.org/abs/1906.02243
WUR (ZD). Precision agriculture - Smart Farming Retrieved from https://www.wur.nl/en/dossiers/file/dossier-precision-agriculture.htm
NYTimes. (2023, June 24). https://www.nytimes.com/2023/06/24/business/ai-generated-explicit-images.html Retrieved from https://www.nytimes.com/2023/06/24/business/ai-generated-explicit-images.html
EU. (2019, April 15). Can artificial intelligence help end fake news? Retrieved from https://ec.europa.eu/research-and-innovation/en/horizon-magazine/can-artificial-intelligence-help-end-fake-news
OPENAI. (ZD.) OpenAI Retrieved from https://openai.com/
IBM. (No Date). AI Ethics Retrieved from https://www.ibm.com/impact/ai-ethics
We will keep you informed by email.
Enter your preferences here.