Artificial intelligence (AI) is changing everybody’s lives in a number of different ways. However, despite the fact that it is progressively being used in more applications, there seems to be a fundamental lack of understanding from the public, policy-makers and even AI developers regarding its capabilities, potential risks and benefits. Moreover, there is not any comprehensive and systematic assessment of the potential (present and future) impact of AI on our society and our planet. Driven by this important gap in the literature, we decided to conduct a thorough investigation to assess the published evidence of positive and negative impacts of AI, and we framed such impacts within the 17 Sustainable Development Goals (SDGs) of the United Nations (UN). These SDGs are subdivided into a total of 169 targets, which define the areas we should be focusing on to ensure a sustainable future, and include Goals ranging from eradicating poverty, to ensuring quality education and equality worldwide, as well as enabling cleaner and safer cities and fighting climate change.
The 17 SDGs span a wide range of knowledge areas, and therefore we decided to put together a group of experts in very diverse topics such as engineering, sustainability, ethics, biology, and of course AI. The study was initiated at KTH Royal Institute of Technology in Stockholm (Sweden), and we involved a total of 10 researchers from all over the world (including the United States, New Zealand and several locations in Europe) in order to provide a comprehensive and unbiased assessment of the literature. We first collected a vast database of studies showing either positive or negative impacts of AI on each of the 169 targets, and then we organized a number of workshops to evaluate each other’s work, discuss, and converge towards a unified view. This process was extremely rewarding and stimulating due to the diverse perspectives from the authors. The main conclusion of the work is that AI can positively contribute to the achievement of 134 targets across all the SDGs, whereas it can inhibit 59 targets. One of the most challenging aspects of this work was to decide what can be considered as “relevant evidence” to establish a connection between AI and a particular target. Indeed, we established different criteria to categorize the types of evidence, based on their appropriateness to establish the connection.
Our research finds that AI can help towards the achievement of all the SDGs, generally through a technological improvement that helps to overcome a certain existing barrier. For instance, AI can help to develop more accurate and robust methods to measure pollution levels in urban areas, thus allowing to develop more comprehensive plans to improve their air quality. Also, satellite images can be analyzed through AI to track and identify areas of growing poverty, with the aim of developing coordinated actions to provide better help. On the other hand, some of the main drawbacks lie in the need for large data centers to perform the required calculations to obtain these results, and the lack of access to such facilities could eventually lead to a net increase in inequalities. Furthermore, we identify that there is a lack of proper regulation towards AI development, mainly due to two reasons: the speed at which this technology is growing, and the general lack of insight into AI. Therefore, we believe that it will be essential to initiate a global debate aimed at developing shared principles and legislation among nations and cultures regarding AI. This will be the only way to ensure a sustainable, AI-facilitated future.
More information here: https://www.nature.com/articles/s41467-019-14108-y