Development practitioners are looking for ways to innovate rapidly how they make decisions and solve problems in the quest to accelerate progress towards the United Nation’s Sustainable Development Goals (SDG) by 2030. Such innovations can be enabled by intelligences, born from innovative uses of new technology. One such promising intelligences is artificial intelligence (AI) – technologies. AI aims at the automation of activities that are typically associated with human thinking, such as learning, decision-making, or natural language processing, and uses algorithms to mimic human learning and cognition, toward addressing narrowly specified tasks.
Artificial intelligence: a strong potential for solving global challenges
The field of AI has accelerated rapidly since around 2010. It has only been eight years since the modern era of deep learning began at the 2012 ImageNet competition. Progress in the AI field since then has been breathtaking and relentless, driven by the growing availability of large data sets being generated in all areas including agriculture, education, human health, commerce, communications, and carried by continued advances in computational power and the development of better machine learning algorithms and techniques.
Today, we have less than 10 years to achieve the SDGs and AI holds great promise. Responsible use of AI can lead to benefits for the development practitioners and beneficiaries For example, the child-growth monitor uses machine learning algorithms to reliably detect malnutrition and further improve their performance as more data is collected. Processing such data and algorithm locally on smartphone, and not on the cloud, could be useful in developing countries and contexts where internet access is limited, and improve privacy protections. Hence, AI can fill gaps in information, identify patterns and relationships, and help foresee risks and opportunities. It can contribute to enhancing knowledge and situational awareness.
Helping development practitioners make decisions and solve problems
New forms of impact assessment and prediction via the emerging uses of AI can also improve impact and identification of optimal interventions. The Red Cross used AI technologies to fuse various data sources to predict overspills of the Nangbeto Dam in Togo. The better forecasts decreased the impact of the overspills and corresponding floods and helped to prepare vulnerable communities.
AI can also open up new forms of smart service provision, humanitarian and environmental interventions, and sources of income. A good example is the startup Apollo Agriculture in Kenya which uses satellite data to train machine learning models that automatically build digital processes for, e.g., customer acquisition or collecting payments. Through these digital processes, Apollo then makes its decisions about lending and credit provision.
Artificial intelligence can expand individuals’ capabilities and skills
Moreover, AI can offer ways to increase individual skills and community capabilities through peer-to-peer knowledge sharing and remote learning. Another example is M-Shule. This mobile platform delivers lessons based on the national curriculum to each student via SMS. It adapts to children based on their skills and abilities using AI technology.
Lastly, emerging uses of AI can increase the agility and efficiency of interventions through automation, freeing human operators for more complex tasks (e.g., analyzing large amounts of data). Machine learning can analyze large amounts of data, which would take a human a long time, and recommend actions based on their analysis. Poor road conditions are a hazard for drivers and a drag on transportation speed, which inhibits economic growth. In Tanzania, an automated road condition survey project evaluated the quality of unpaved roads using satellite images and deep learning techniques. The algorithm managed to assign images to road conditions with 73% accuracy, flagging images for which human judgement was required.
But the uses of artificial intelligence are not without risks. Responsible use of AI can help achieve these values, which are important to development efforts, but development practitioners must be careful. They should monitor for emerging negative externalities, and challenges and consider whether they can be mitigated and at what cost.
Mitigating the risks associated with uses of emerging technologies
Given that AI rely on quality training data; gaining access to data does sometimes entail investment in data collection technologies or forging partnerships with data-holding or data-generating collaborators. Lack of quality training data in term of diversity and representation can lead to bias, discrimination, and erroneous conclusions that will not be useful in achieving development objectives. In addition, the failure to incorporate ethical requirements within machine learning algorithms can also contribute to potential negative impacts of development projects.
Although AI methods can increase efficiency of, for example, manufacturing companies, it may result in reduction of low-skilled labor. Development practitioners can be key in efforts to ensure that the transition toward an increased adoption of AI methods happens in an inclusive manner.
Data privacy, one of the overarching challenges of the digital era
Moreover, data privacy issues play a significant and often limiting role in AI’s trajectory, as data is the lifeblood of AI. The General Data Protection Regulation (GDPR), introduced by the European Union (EU), can hinder potential positive impacts of AI for development if data protection principles are not well considered. The development of AI systems that are set to be privacy by design and by default often entail some additional costs. The involvement of data protection authorities in data assessment can help mitigate any such challenge. It is worth mentioning that methods that enable AI models to learn from datasets without compromising their privacy is thus becoming an increasingly important pursuit.
Development practitioners should actively engage in a dialogue with data protection authorities, and all stakeholders, including controllers, and civil society, in order to develop appropriate responses, based on shared values and effective technologies. There is a need to promote a broad societal debate on AI applications to provide high-level solutions. Responsible use of AI with consistent application of data protection principles can contribute to the success of AI applications in development, by generating trust and preventing risks.
Given the fast pace at which the field of artificial intelligence is moving, five years from now, models that are currently considered cutting-edge will have become outdated. The next generation of artificial intelligence, with novel AI approaches like unsupervised learning, Federated Learning and Transformers, will unlock currently unimaginable possibilities in technology and hence for development.
This article is published in conjunction with The Conversation.