To prevent climatic phenomena extreme, optimize the use of water for agricultural purposes, use the Facial recognition to detect child malnutrition, combat early school leaving or identify the most intolerant places: these are just some of the cases of use of artificial intelligence for a good purpose, collected in the volume Artificial Intelligence for Sustainable Development.
It is one rich and articulated study, carried out by Ong 2.0, Cisv, AIxIA (the Italian Association for Artificial Intelligence) and from Department of Computer Science of the University of Bari Aldo Moro, with the support of the Foreign Ministry, 4 Italian companies in the sector and Cnr. A long research work that has resulted in one of the most complete publications on the subject, at an Italian level. We start from a broad definition of AI: “Artificial intelligence is a discipline, belonging to the field of computer science, which studies methods and tools that allow a machine to do things that appear intelligent”. We try to explain a complex phenomenon in simple terms, providing clear definitions and possible developments in the more or less immediate future. Will we have to compete with a weak AI, which will allow machines to behave as if they are intelligent, or with a strong AI, thanks to which the machines will be able to think?
Above all, in the introductory phase, in addition to telling the story of AI, photographing its current state and possible developments, the volume also raises uncomfortable questions, which in a path such as that of technological innovation, which sometimes leads to a a priori enthusiasm and at times almost fideistic, are not so obvious. One above all: “The question we must ask ourselves is whether it is really worth building such machines”. How to deal with the so-called technological singularity, that moment in which progress will accelerate beyond the capacity for forecasting and understanding of human beings? How can this moment be managed? They are not missing fears related to the development of AI, but perhaps “we should also be afraid of a world where we are unable to overcome the limits of our present intelligence, which appears inadequate to the challenges we face”.
Artificial intelligences between racism and the ethical question
by Emanuele Capone
Artificial intelligence for sustainable development
Among these challenges are the 17 Sustainable Development Goals identified by the United Nations. And the central part of the volume focuses on these. For each goal it is described how artificial intelligence systems can be used to facilitate its achievement. The pros and cons are highlighted and above all concrete cases of application already in progress are reported.
As part of the fight against poverty (first of the objectives), in 2020, a Nature Communication article highlighted the use of deep learning to measure the level of well-being of African villages. Using publicly available multispectral satellite images, deep learning models were trained to predict estimates based on a survey of the wealth of some 20,000 African villages. The method takes as a reference the light emissions of the villages, correlating the issue to wealth. The estimate has some margin of imprecision, but not higher than those present in the measurements carried out by census, in a complex territory such as Africa. Satellite estimates proved very useful and allowed for the creation of a map of wealth.
To combat hunger, two best practices stand out from the others: in Chile, NotCo leverages AI to create new sustainable food products, creating formulas that perfectly reproduce the flavor and texture of foods, without using foods of animal origin. NotCo sells milk that does not come from cows and mayonnaise that is not made with eggs and to do so, it uses a supercomputer equipped with artificial intelligence, which is responsible for combining the ingredients.
Kimetrica, social enterprise based in the United States, Ethiopia and Kenya, through Meron (the acronym stands for Method for extremely rapid Observation of nutrional status) manages to detect malnutrition simply from photographs. Thanks to facial recognition technology it is possible identify malnutrition in children during health emergencies.
Regarding the protection of water resources, the case of ConserWater is very interesting, an Indian water conservation platform that uses satellite data and artificial intelligence to accurately calculate the water needs of crops throughout India, helping farmers to reduce water waste and to distribute water resources adequately and efficiently among different types of crops.
social and security
Behind the scenes of TikTok, where AI and humans fight against fake news
by Emanuele Capone
The consumption of artificial intelligence
Artificial intelligence algorithms, therefore, can help in achieving sustainability objectives defined by the 2030 Agenda. But there is also another side of the coin: i University of Massachusetts Amherst researchers, carrying out an assessment of the energy needed to train AI, found that 284 tons of carbon dioxide equivalent can be emitted, emissions equal to almost 5 times those of the average life of an American car, including production.
Deep neural networks need a large processing power and significant amounts of energy, often generated by fossil fuels, with a negative impact on the environment. A challenge within a challenge. But maybe it will be an artificial intelligence to help us to find a solution.