Artificial intelligence (AI) is already embedded in a range of digital services. Voice assistants such as Alexa, car routing or content translation all involve machine learning - the most popular form of artificial intelligence technology. There are many warnings these days about AI, such as the ethics behind these machine driven decision systems or threats of automation and the loss of many jobs.
Very little is reported about how artificial intelligence can improve public services and can have positive social impact. Smart algorithms combined with cloud computing power allow unprecedented forms of data analysis that would take much longer if humans were doing it.
Below I researched a list of inspiring examples of how artificial intelligence is used in public service, education, human rights and health (#aiforgood). The examples prove that AI can have a helpful impact, but as any other technology, does have consequences. The case of depression detection shows the challenges of such approaches. At the end of the day it is still an algorithm that can lead to false predictions. So it is very important to weigh the risks of false decisions in each of these projects.
Inspiring examples of artificial intelligence for good (#ai4good)
Health: Snakebite is the second most deadly neglected tropical disease. Snapp, a first medical decision‐support tool for snake identification based on artificial intelligence, can help to save lives. With the mobile app, patients make a photo of the snake bite that identifies the type of snake and react accordingly. AI is used to analyze thousands of snake bite photos to provide the best possible accuracy with a new snake bite. Conclusion: Great combination of artificial intelligence and a real world problem.
Health: Corti listens in on emergency calls and analyzes conversations in real time. Utilizing machine learning, Corti helps dispatchers diagnose illnesses and provides prompts for effective action. “After analyzing data from Corti’s first trial run in Copenhagen — which included 4,089 emergency calls regarding cardiac arrests — it was discovered that human dispatchers managed to recognize 73.9 percent of cardiac arrest calls while the Corti AI correctly analyzed 95.3 percent of the calls.” TNW Conclusion: Quite a sophisticated approach using different AI methods and promising for human-assisted artificial intelligence.
Health: Artificial intelligence that helps to detect depression. Researchers at the Massachusetts Institute of Technology (MIT) used deep learning technology in an experiment to predict if a person has a mental illness. A chatbot asks a person various questions to give an 85% accurate prediction. Conclusion: It is incredible and scary how behaviour and even social status can be analyzed thanks to facial recognition and human-machine interaction. One needs to be cautious that such services are implemented for the benefit of the person.
Health: Ada, a personal mobile health guide invented in Germany, is an intelligent chatbot that interviews people who may feel a pain or have a health related issue and want to find out the source of the problem. In a step-by-step interview the chatbot attempts to diagnose the potential causes of the pain. Ada acts as a first diagnosis agent that helps people solve an issue. Conclusion: Great scaling concept with worldwide users and particularly helpful in contexts where doctors are not within reach.
Public service: Community radio is an important communication channel in African countries. The United Nations Global Pulse project records all these radio channels in Uganda in order to track public conversations. Analyzing these radio streams can help to detect crises early to avoid disasters such as famine. Conclusion: Innovative idea to track radio and with great potential if the algorithm really detects such phenomena.
Public Service: Pothole detection thanks to smartphone cameras and machine learning by the German startup Vialytics. The next step of the fix-my-street concept. Cars from the local municipality can be easily equipped with smart phones to film the road. Such photos can be then analyzed to identify the types of potholes and predict where the next potholes could potentially appear. Conclusion: Interesting expandable idea, but still not really disruptive, because people have to spot the holes and make the images.
Public Service: An estimated 4 billion people in the world lack a physical address. Without an address, citizens lack access to important civil services or medical care. A MIT project attempts to use machine learning to analyze satellite data to construct an address system. It is a similar attempt like What 3 Words, whereas the latter needs no AI and already has companies adopting it. Conclusion: Interesting approach that shows how much can be done combining AI and satellite data, but lacks a concept of implementing it.
Education: A key challenge for policy makers is the lack of reliable and up-to-date statistics. That applies also for the education sector, where not all countries provide reliable data. Using satellite data to identify school buildings worldwide can help to infer education statistics (e.g. number of students, distribution of schools). The first experiments show an accuracy of 75% to find schools. Conclusion: Satellite data offers a range of great opportunities to use artificial intelligence and can improve statistics worldwide in many areas.
Education: Squirrel, a personalized learning platform that provides customized teaching for each student using artificial intelligence. Learning levels are measured and exercised in real-time to provide students with optimal learning content and speed. Next to paid services, Squirrel also offers free education to thousands of children in over 1000 learning centers across China. Conclusion: Great potential for effective personal learning, but not affordable for all parts of society.
Human rights: Human trafficking is a big challenge in India. A quarter of a million kids were called missed in the past 5 years. The Indian project Track Child uses the latest face recognition technology to compare photos from missing kids with photos from kids from orphanages. Within a few days, 2930 kids were identified. Conclusion: One of the cases where face recognition really makes sense in finding missing kids.
Human rights: Amnesty International uses satellite images to detect human rights violations, for example in Darfur, Sudan. Through the “Decode the Difference” project, volunteers analyzed imagery for example for destroyed buildings. Machine learning can help here to analyze more amounts of images faster. Humans are required to establish a first database of types of house destructions and then machines take over to potentially analyze satellite data in real-time. Conclusion: Interesting human machine cooperation, which shows again the value of satellite data and the potential of AI for good in human rights.
Environment: Wildlife conservation involves counting animals, which is a difficult task. Motion-sensitive cameras can automatically photograph wild animals, providing massive amounts of data. In the Snapshot Serengeti project scientists used deep learning to analyze these images. The system can save 99.3% of the manual labor ( more than 17,000 hours) while performing at the same 96.6% accuracy level of human volunteers. Conclusion: Great example where AI can save a lot of resources and time.
These examples show how artificial intelligence for good is already used in different sectors by many organizations. They are not necessarily intelligent systems, but are only reliable if they were trained previously by humans and lots of data. But particularly in the non-profit sector there is little open data available to work on such innovative services. And some services would include personal data and should not be implemented at all, because the benefits are not worth the loss of privacy.