Ricardo Vinuesa, KTH Royal Institute of Technology, highlighted that AI can have a beneficial impact on 79% of the SDG targets, inhibiting 35%.
Among AI’s numerous positive applications for the SDGs, he highlighted matching electricity supply and demand, making better climate models, identifying areas of high concentration of plastic pollution, and “finetuning” school curricula to individual students
The International Telecommunication Union (ITU) and 40 UN partners convened an event on ‘The Role of AI in Achieving the Sustainable Development Goals,’ with a specific focus on SDGs 11 (sustainable cities and communities) and 13 (climate action). Speakers explored AI’s potential in revealing hidden connections among the SDGs to optimize policy responses for better solutions.
Co-convened with the Government of Switzerland on 2 February 2023, the event was first in the AI for Good Discovery series.
According to the ITU event page, while AI can enable 134 of the 169 SDG targets across all the Goals, it may inhibit 59. Regulatory insight and oversight for AI-based technologies, the introduction notes, need to support the fast development of AI in a way that would enable sustainable development. It warns that failure to do so could result in transparency, safety, and ethical gaps, leading to suboptimal outcomes for the SDGs.
Markus Reichstein, Max Planck Institute for Biogeochemistry, moderated the discussion.
Ricardo Vinuesa, KTH Royal Institute of Technology, presented research by an international consortium on how machine learning could support the SDGs. He said a multi-disciplinary team, spanning the areas of theoretical and applied machine learning, biodiversity, sustainable energy, law, and ethics, among others, conducted a “consensus-based expert elicitation process” to identify published evidence of AI acting either as an enabler or as an inhibitor of the Goals and targets along the environmental, social, and economic dimensions.
Vinuesa highlighted that AI was found to have a beneficial impact on 79% of the SDG targets, inhibiting 35%, with environmental targets attracting the most positive contributions, social targets – the least, and economic targets requiring further investigation in terms of potential positive effects of AI.
Among other key results he noted that AI enables new technology applications, such as using satellite data to track poverty, and that gender gaps in data and AI workforce both need to be tackled to enable SDG 5 (gender equality).
Vinuesa further highlighted that:
- Current electricity use by the information and communication technology (ICT) sector stands at 1% of total use – a figure that is expected to grow to 20% by 2030, revealing AI’s impacts on the achievement of SDGs 7 (affordable and clean energy) and 13;
- All SDG 11 targets can be positively affected by AI, and machine learning can be used to reproduce, measure, and predict, with high levels of accuracy, pollution in cities and specific urban areas;
- Data-driven COVID-19 contact tracing apps would have had more impact on SDGs 3 (good health and well-being) and 10 (reduced inequalities) had there been more penetration; and
- While AI technology is developing very fast, its uptake by individuals is lagging behind, with the slowest change observed for governments.
Among challenges to maximizing AI’s enabling potential for the SDGs, Vinuesa mentioned the lack of interpretability of deep learning models. Highlighting the need to exercise caution when assessing the impact of AI on the SDGs due to the Goals’ interconnectedness, he cited an ongoing study that uses natural language processing to identify synergies and trade-offs among the SDGs humans may not pick up on, with the objective of developing an algorithm to help avoid unexpected negative interactions among the SDGs in policy decisions.
Participants then raised: the principles of transparency and interpretability as essential parameters for AI ethics; the need to make AI itself more sustainable by, for example, developing more efficient algorithms; and the need to increase the capacity of developing countries to build their own AI models.
Among AI’s numerous positive applications for the SDGs, Vinuesa highlighted matching electricity supply and demand, making better climate models, identifying areas of high concentration of plastic pollution, and “finetuning” school curricula to individual students. [Event ITU Page] [SDG Knowledge Hub Sources]