Mar 12, 2018 | Author: Patricia Morén / Jordi Suñé
Artificial intelligence (AI) increases exponentially, but does it do it in the same way worldwide? The answer is No. While the richest areas of our planet and the most wealthiest social classes have access to each new solution and service of AI systems, the poorest run the risk of not accessing this force of change in science and the world economy. Conscious of this reality, the next AI Summit in Geneva will focus on the use of AI to help achieving the sustainable development goals of the 2030 Agenda, the first of which is the end of poverty.
The British physicist Stephen Hawking, recently deceased, said that the complete development of artificial intelligence (AI) could lead to the end of the human species and argued that the priority research lines should not only focus on improving AI capabilities, but also "to maximize the social benefit of AI". Hawking stated this in Research Priorities for Robust and Beneficial Artificial Intelligence: An Open Letter that he made public in 2015. Since then, 8,000 scientists from all around the world have signed it. Hawking predicted that the social impact of AI would only continue to grow and that with the AI "the eradication of poverty and disease is not unfathomable". But also that, due to its enormous potential, "it is important to research how to reap its benefits while avoiding pitfalls," he warned in the Open Letter.
The focus: Sustainable Development Goals
The same year, in 2015, the United Nations (UN) adopted the 2030 Agenda on Sustainable Development, with 17 Sustainable Development Goals (SDGs), to which signatory countries and societies should aspire for that by 2030. Among these SDGs there are included many goals directly related to underdevelopment and poverty, such as the end of poverty, zero hunger, good health and well-being, gender equality, clean water and sanitation, decent work and economic growth, climate action or reduced inequalities, among others.
Today, in line with that universal call of the SDGs and the Hawking's Open Letter, the next AI for Good Global Summit, to be held in Geneva (Switzerland) from 15th to 17th of May, 2018, has the vocation to seek AI solutions that can have a beneficial long- term impact to achieve the Sustainable Development Goals. This second Global AI Summit is being held under the slogan "Accelerating progress towards sustainable development goals" and recapturing the spirit of the first and successful Global AI Summit for the Good of 2017, which also took place in Geneva. Both important meetings have been organized by the International Telecommunications Union (ITU), the United Nations specialized agency for Information and Communication Technologies (TIC), in partnership with the XPrize Foundation, the leading foundation in incentivized prize competitions, based in Culver City (California).
The editions of the AI for Good Global Summit is the leading platform of the United Nations for dialogue on artificial intelligence. The new edition of 2018 is action-oriented and, therefore, aims to identify practical AI applications and support strategies to improve the quality and sustainability of life on our planet. Among other experiences, innovations on the potential of AI to map poverty and aid from natural disasters will be shared and presented, using satellite imagery or how AI could help provide citizen- centered services in smart cities, to achieve universal health coverage and, finally, to help transparency and explainability in AI algorithms.
Pilot experiences of AI against poverty
There are around 3 billion people in the world who still live in poverty. More than 1 billion people have emerged from extreme poverty since 1990, according to the United Nations Development Program, but 702 million people in the world - 9.6% of the world's population – still living in extreme poverty, which means that they live on less than US $ 1.90, according to the latest data available from the World Bank.
Despite these estimations of the population affected by poverty, it is not well known where the poorest people live exactly, since household surveys are almost not done in the most disadvantaged and distant areas of the planet, according to Marshall Burke, an economist at Stanford University, in California, in an article in Nature, where the author, Declan Butler, collects some of the most interesting AI pilot experiences related to poverty.
One of them is the group of Neal Jean, Burke and other researchers of Stanford, who are training algorithms in which they use satellite imagery at night, where the illuminated areas serve as an approximate indicator of prosperity, so that they can deduce the characteristics of satellite diurnal images, such as roads or roof types, that correlate with relative wealth or poverty. In particular, in a pilot study with five African countries (Nigeria, Tanzania, Uganda, Rwanda and Malawi), published in Science, this team of researchers has discovered that their AI system is more accurate than previous methods that only used night lights, when assessing the wealth of a village. In particular, the researchers show how a convolutional neural network can be trained to identify image features that can explain up to 75% of the variation in local-level economic outcomes.
There are another initiatives to study the spatial distribution of poverty. For example, the researchers at Carnegie Melton University use a platform, the AI Penny system, that allows to understand how wealth and poverty of the cities are viewed from the space, through an artificial intelligence system based on automatic learning by using neural networks. This tool enable to play with the landscape of a city and to add or to eliminate different features of buildings, parks or highways in high resolution satellite images, from DigitalGlobe. The researchers have entered into the system data on the average family income levels of the districts of two US cities, New York and St. Louis. Afterwards, they have colored the different areas according to income level and have been able to draw up a family income map in the city, where green represented the areas with the greatest purchasing power, red was the lowest annual income and orange and yellow, the average levels.
The initiatives aimed at geolocating poverty are not the only ones to fight against it. Other projects focus on combating other faces of poverty. As explains Butler in the above mentioned article in Nature: This is the case of another Stanford team led by Jiaxuan You, which uses AI and satellite remote sensing data to predict crop yields a few months before harvest, in order to anticipate food shortages, or the case of UNICEF, the UN charity for children, which is evaluating whether Deep Learning can help diagnose child malnutrition from photographs and videos of children at risk, compared to the manual method currently used, which consists of measuring the circumference of the middle arm, that is not always accurate.
AI companies for social benefit
These are some of the ideas put in place to face the most pressing challenges of our world. The XPrize Foundation, co-organizer of the Summit, has announced important funds to finance ideas and the industry has joined this stream of good to research in AI for social purposes: from OpenAI, a non-profit research company founded in 2015, until the Partnership on AI founded in 2016 by the giants Google, Amazon, Apple, DeepMind, Facebook, Microsoft and IBM; the latter has launched the program known as Social Science for Good.
At the moment, AI seems to be under control, at the service of achieving Sustainable Development Goals. But we must not lower our guard or lose the sight of Hawking's message in his Open Letter: "We can not predict what we could achieve when this intelligence is magnified with the tools that artificial intelligence can provide", so science should refrain from creating something that humanity can not control. It is necessary to act with human intelligence so that artificial intelligence continues to be at the service of Good and Poverty.