Applying Artificial Intelligence to Solve Development Challenges: The Potential and Pitfalls

RTAC

  • POSTED June 26, 2019
  • General Information

By Michael Shoag, Vice President Government Services for Forum One, a partner on USAID’s Research Technical Assistance Center.

Artificial intelligence (AI) technology is growing and evolving daily. In May, USAID’s Research Technical Assistance Center brought together more than 80 experts in IA and international development to discuss digital tools and the future of international development. This meeting was an opportunity to discuss AI, its potential, and the challenges that need to be managed when applying AI in development initiatives.

The Potential

AI refers to digital tools that learn from experience, adjust to new information, and are able to perform tasks that normally require human intelligence. AI technology can help developing countries in many ways. Three specific types of AI and their potential applications follow.

Machine Vision. While computers can’t see, they can analyze images to find features, categorize data, or identify differences. Machine vision can be used in development to look at images from satellites, drones, TV cameras or other sources. Machine vision can help people:

  • Map slums.
  • Identify which crops are being planted where.
  • Show location of crop disease or insect infestations.
  • “Read” X-rays, MRIs and other medical images to find health issues.
  • Understand the scale of a disaster.
  • Estimate populations.

Natural Language Processing (NLP) is a type of AI that focuses on understanding written or spoken language. NLP can be tricky to use in developing countries. While it works well in English and the most popular languages, it has not been highly developed for the vast majority of the world’s more than 7,000 languages. Some uses of NLP include:

  • Automatic translations.
  • Helping low-literacy individuals or communities access information.
  • Better meeting the needs of those with physical disabilities.
  • Answering questions with chatbots, which can hold a conversation.
  • Examining documents to find patterns or relationships.

Networking Analysis is a form of AI that looks at networks to understand relationships and uncover information that is difficult to otherwise see. Network analysis can be used in developing (and developed) countries to:

  • Identify human trafficking.
  • Predict animal poaching areas.
  • Understand terrorism networks.
  • Identify fake news.
  • Identify false actors.
  • Understand social relationships.
  • Determine how diseases will spread.
  • Map how information flows.
  • Find relationships between health and social relationships.
  • Uncover tax fraud.

The Pitfalls

While the potential for AI is clear, it’s important to proceed with caution, given the possible complications. Several challenges to applying AI—which may be magnified in developing country contexts—are outlined below:

  • There tends to be less data and less clean data in developing countries. Clean, well-labeled data is needed to “train” AI. In developing countries, the data sought may not exist.
  • Many developing countries don’t have the institutional capacity to find, contract, or use AI effectively.
  • Developing country government officials may be skeptical of the benefits of AI. They may want to see evidence that there is a marked improvement from using AI before committing to using it in their country.
  • AI related to language supports only the most popular languages. Billions of people in developing countries can’t use AI focused on Natural Language Processing. This tends to impact the poorest, whose native language or dialect is not supported. They can’t use AI to translate, to ask questions, or to overcome illiteracy through talking and listening, as opposed to reading and writing.
  • Preserving privacy represents another challenge for AI, which can be a factor when looking at personal health data, images or facial recognition, social networking data, voice data, or other forms of data.
  • If weaponized against citizens, AI can be dangerous. Just as IA can help identify terrorist networks, it can also be used to find networks of those who don’t support a politician or government policy, to follow and track individuals, and to restrict communication in whole, or in part, on social networks.

 

What’s next. We can expect to see huge advances in the use of artificial intelligence in everything from medicine to transportation to agriculture in the coming years. These may help low-income countries to improve many aspects of life. But first, robust ethical standards and strategic solutions will be needed to address technical challenges and ethical considerations associated with AI in the developing world.