Friends or foes: A trade-off analysis of artificial intelligence (AI) and sustainable development

Artificial intelligence (AI) can enhance and accelerate efforts towards achieving the Sustainable Development Goals (SDGs), but it also has the potential to jeopardise progress, adversely impacting 35% of the SDG targets.

Artificial intelligence and sustainable development

Artificial intelligence (AI) has considerable potential to positively contribute to the three pillars of sustainable development, namely economy, society, and the environment (e.g., Galaz et al., 2021; Korwatanasakul and Takemoto, 2021; Nishant, Kennedy, and Corbett, 2020). In general, for all stakeholders, AI applications can help to design, plan, execute, and evaluate programmes that enhance progress on the SDGs through increasing their efficiency and effectiveness.

The manufacturing sector utilises AI and associated technologies to manage non-renewable resources and supply chains, diffuse knowledge and expertise, build effective multi-sector partnerships, and improve business practices towards sustainability. For instance, in Kenya, the startup Apollo Agriculture deployed machine learning in customer acquisition management and digital payments, expanding financial inclusion to a broader range of customers (Addo, 2021).

Furthermore, AI applications offer a range of new possibilities for the public sector to make informed decisions and optimal interventions to solve economic, social, and environmental problems. For example, AI technologies predicted overspills of the Nangbeto Dam in Togo, helping vulnerable communities prepare for floods (Addo, 2021).

In addition to SDGs-enhancing programmes, AI directly and positively affects 79% of SDG targets, including 93% of the environmental targets, 70% of the economic targets, and 82% of the social targets (Vinuesa et al., 2020). AI applications — such as climate modelling, smart fishery conservation, land-use monitoring, and smart grids — generate insights on environmental issues (particularly SDGs 13, 14, and 15) and thereby monitor and control changes in activities affecting the environment, e.g., marine life migration, fishing activities, pollution levels, and disease vectors and outbreaks. The International Energy Agency (2021) reports that digital technologies, including AI applications, are driving sustainable energy solutions that enable cities to realise net-zero emissions through smart control of buildings and transport systems (SDGs 7, 11, and 13).

AI technologies, such as precision medicine, e-health, autonomous vehicles, smart farming, business intelligence, automation and robotics, can advance economic and social targets. Through digital platforms, AI creates new business and job opportunities, boosting the economy even during the COVID-19 crisis (SDGs 1, 8, and 9). The International Telecommunication Union (2018) forecasts that by 2030 AI could potentially contribute approximately USD 13 trillion to the global economy, with 1.2% annual growth. It can also help alleviate poverty and hunger (SDGs 1 and 2) through data analytics techniques that map poverty and food security while improving harvest yield and food safety.

Challenges

Despite the benefits of AI, the adoption rate of AI technologies is low and is increasing very slowly in both public and private sectors. AI technologies require intensive investment and their economic impact is gradual. There is also a lack of capacity in terms of human capital and research to understand complex AI applications (How et al., 2020; Mehmood et al., 2020). Vinuesa et al. (2020) also found that AI potentially jeopardises progress on the SDGs, with an adverse impact on 35% of the SDG targets (particularly SDGs 1, 4, 6, 9, and 10), highlighting the risks of unregulated AI (Truby, 2020).

AI negatively impacts economic development due to the digital divide and, in turn, unequal distribution of the benefits of the technology. AI potentially worsens income inequality at the individual, business, and national levels as the benefits only accrue to those who have access to AI technologies. Moreover, automation-driven technologies are replacing routine jobs faster than ever, such as seeding in the agricultural sector, assembling parts and components in the manufacturing sector, and serving and handling payments and receipts in the service sector.

AI-associated risks also result in social problems, e.g., cybercrime and discrimination. Korwatanasakul and Takemoto (2021) highlight that Japan, one of the most advanced economies, is still in the early stages of AI policy development, focusing on general discussions on cybersecurity, ethics, governance, and rules and regulations — which is indicative of the broader lack of discussion on these topics in most parts of the world. The advancement of AI requires a large amount of data while compromising privacy concerns and security. In addition, recent incidents, such as Amazon’s sexist AI recruiting tool (Dastin, 2018) and the racist AI algorithm of Facebook and Google (Vincent, 2018; Angwin and Grassegger, 2017), pose significant concerns regarding discrimination, resulting in economic and social disadvantages.

AI applications, especially machine learning, contribute to a high carbon footprint and intense energy consumption due to the computational resources and large data centres needed. For instance, training OpenAI’s giant GPT-3 text-generating model consumes energy equivalent to a round-trip from the earth to the moon (Quach, 2020).

Ways forwards

Long-term strategic planning with an agile decision-making process: Policymakers should adopt long-term strategic plans for AI development while accelerating the overall development of AI policies. A long-term strategic plan helps communicate with stakeholders the long-term benefits of AI and promotes risk-free and human-centric AI, which is needed to ensure equality and protect people’s human rights and (cyber)security when planning and mapping the ecosystem of the AI sector (Korwatanasakul and Takemoto, 2021).

Inclusive AI: As the disproportionate distribution of AI benefits exist at the national, corporate, and individual levels, policymakers should involve all stakeholders through supportive programmes, such as financial inclusion and capacity building. Moreover, forging public–private partnerships at the domestic and international levels potentially helps promote more inclusive AI through the diffusion of knowledge and technologies from the global North to the global South. Knowledge sharing initiatives, such as EU-Japan.AI (https://project.eu-japan.ai/), provide platforms for funders, researchers, investors, and policymakers to exchange information internally and expand their networks to involve more stakeholders. Inclusiveness also involves the development of unbiased AI, which offers informed decisions and solutions towards all groups of stakeholders, regardless of sex, race, age, and other factors. Policymakers should conduct comprehensive assessments of social, economic, and cultural factors for AI training/learning to prevent biased AI (Mehmood et al., 2020).

Minimising the trade-offs between AI advancement and sustainable development: It is important to understand the trade-offs between AI advancement and sustainable development. Thus, careful impact assessments of AI technologies are essential in every stage of development, especially pre-assessment. In addition, policies with interventions specific to each SDG target and context are also necessary to minimise adverse risks and consequences for the economy, society, and environment.