Algorithms of applications and social networks, autonomous vehicles, automatic translation, facial recognition, machine learning, artificial neural networks, medical diagnosis... There are many concepts, technologies, and uses of artificial intelligence (AI) increasingly present in everyday life, a result of the great development of the field in the last decades.
In order to provide non-specialized audiences with a comprehensive view of this technological revolution, the 101st edition of the journal Estudos Avançados devotes its main dossier to the discussion of the current state, perspectives, and impacts of AI. The digital version (Portuguese only) is available for free at SciELO.
Composed of nine articles authored by 17 researchers from USP, University of Campinas (UNICAMP), Federal University of Rio Grande do Sul (UFRGS), and Federal University of Pernambuco (UFPE), the dossier "Artificial Intelligence" analyzes the development of the field since its origin in the 1950s, its numerous applications, and the debates it raises in the scientific and technological scenario. They do not neglect "the risks, the associated ethical care that its massive employment requires, and the social, political, cultural, and moral implications that rapidly transform contemporary societies," as summarized by the publication's editor, sociologist Sérgio Adorno.
The opening of the dossier is dedicated to the methodological aspects of AI research, with an article by Fabio Gagliardi Cozman, from USP's Polytechnic School (EP). He explains that there are two "fundamentally different styles of approach to AI: on the one hand, the empirical style, strongly supported by observations about the biology and psychology of living beings and ready to embrace complicated architectures that emerge from the interaction of many disparate modules; on the other hand, an analytical style supported by general and organizing principles, interested in abstract conceptions of intelligence and assisted by mathematical and logical arguments."
According to Cozman, around 1980, the terms "scruffy" and "neat" were coined to respectively refer to these two styles of work in AI. This methodological divergence remains with the constant dilemma between "the search for rational artifacts based on clear principles or empirical artifacts that reproduce patterns," he says. His proposal is to invest in architectures based on principles of rationality that allow to simultaneously house several modules, many of which are based on massive data collection.
The concern with the form of AI development is shared by André Carlos Ponce de Leon Ferreira de Carvalho, from USP's Institute of Mathematics and Computer Science (ICMC): "What we have to decide now is no longer whether or not we will have AI, but how we will have it." To reduce possible risks "it is necessary to develop new AI algorithms or to use them in new and innovative ways, taking ethical, social, and legal issues into account," he points out.
Reasons for excitement
This new period of euphoria in relation to the possible benefits of using AI is due to three factors, according to Jaime Simão Sichman, from EP: the current low cost of processing and memory, the emergence of new paradigms such as deep neural networks, and the huge amount of data available on the internet due to the large use of resources such as networks and social media.
Sichman warns that there are potential risks in this "technology, as in any other, which can be provoked if the actors involved in the production and regulation of its use do not create an adequate space for discussing these issues."
According to Teresa Bernarda Ludermir, from UFPE's Computer Center, the "extraordinary" advance of AI in recent years and its importance in solving technological and economic problems is mainly due to machine learning techniques, especially the use of neural networks. In addition to addressing the current status of the area, and its research challenges and opportunities, her article mentions social impacts and ethical issues arising from the uses of AI.
The changes that have occurred in AI since its emergence, especially regarding educational systems, are the object of the panorama presented by Rosa Maria Vicari, from UFRGS's Institute of Informatics. The researcher recalls that in 1980 and 1993 "the applications were interesting, but did not provide adequate answers in terms of language comprehension and medical diagnosis." In the last two decades, however, "the applications have remained, but there have been advancements in automatic translation, image recognition, cancer diagnosis. and autonomous cars."
Some applications
Nine researchers from UNICAMP's Institute of Computing are the authors of an article on digital forensic science, which is the use of scientific methods and techniques for investigating crimes in the digital world. According to them, the current importance reflects from the challenges resulting from the emergence of social media and the immense volume of data that they generate, intensified by the advances of AI. Such amount of data is analyzed using AI techniques. The article presents challenges and opportunities associated with the application of these techniques, and examples of their use in real situations.
A specific case in which AI plays a fundamental role is natural language processing (NLP), which is essential for analyzing large amounts of data contained in texts, among other uses. The topic is discussed by Marcelo Finger, from USP's Institute of Mathematics and Statistics (IME), in particular regarding the Portuguese language. He explains that the NLP is at the confluence of areas such as computer science, linguistics, logic, psychology, among others, and requires a multidisciplinary approach by nature.
However, all the advancements in NLP with the consequent generation of products and the facilitation of a series of services "seem to have brought no substantial information about the human process of reproducing and communicating through language," says the researcher. Following this line of reasoning, "natural language processing would have dissociated itself from language study." There are those who say that "technology will eventually kill the traditional study of language," he adds. For Finger, these two views are exaggerated: "Linguistics is absolutely fundamental for the area of language processing, since this computational task does not explain the language, and does not help to predict or explain the natural evolution of languages."
Ana Bazzan, from UFRGS's Informatics Institute, analyzes the use of AI to improve transportation systems. For her, the application of AI in the service can improve the use of the existing infrastructure in order to better meet the demand of displacement of people and goods. Her article addresses two tasks in which AI has relevant contributions in the sector: the control of traffic lights and the choice of routes.
Impacts on labour
The dossier ends with an article by Ricardo Abramovay, a senior professor at USP's Institute of Energy and Environment (IEE), on one of the issues that most concern society with regard to AI: the loss of job positions. The author states that, although the most advanced forms of the digital revolution (AI, machine learning, and the internet of things) are replacing a considerable part of the workforce, "this is not where its greatest threat lies." The problem is that this revolution "is strengthening a social polarization of the labour market that goes against the foundations of the welfare state in the 20th century," he says.
According to Abramovay, "the place of work in the cohesion of contemporary societies involves a fundamental philosophical discussion: what work is, what employment is, but more than that, how we can make our capacity for cooperation result in a better life for everyone in ways that are not unworthy and undervalued for the overwhelming majority, alongside creative and uplifting activities for a small minority."
Other sections
In addition to the main dossier, the issue includes an encouraging study on the diversity of higher education institutions in Brazil, a set of texts on urban agriculture, and reviews of five editorial releases.
In terms of pursued objectives and obtained results, this diversity requires an effort to classify the institutions by similarity of profile, establishing different evaluation criteria for each group. This is what the article "For a Typology of Brazilian Higher Education: Test of Conception," by Simon Schwartzman, Roberto Lobo Silva Filho, and Rooney Coelho, proposes.
According to the authors, the differences between the institutions are not recognized with all the implications by the legislation or by the evaluation system adopted by the Ministry of Education. The article presents a typology proposal that seeks to clearly identify these differences in order to serve as a basis for an information system and evaluation procedures.
To this end, the researchers propose to group institutions with similar profiles from the point of view of their size, legal nature, and involvement in teaching and postgraduate activities, in addition to verifying the extent to which this differentiation corresponds with the diversity of characteristics of professors, students, and their segment. The final part of the study discusses some of the implications of the typology for the higher education assessment system and for improving the quality and performance of higher education in the country.
Urban agriculture
According to the editor of Estudos Avançados, the articles on "Urban Agriculture" address specific issues connected with each other as modalities and alternatives for the promotion of food security.
The six texts discuss the multifunctionality, production, and commercialization of urban agriculture, its association with agroecology, the importance of community gardens and backyard gardens, and the contradictions of locavorism (food activism that emerged in the past decade and privileges the consumption of locally produced food) in the face of the experiences of urban agriculture in São Paulo.
Reviews
A set of six reviews addresses works with varied themes, including geology, history, intellectual production, and literature. Ricardo Soares and Wilson Machado write about "The Anthropocene as a Geological Time Unit: A Guide to the Scientific Evidence and Current Debate," which gathers data from the 35th International Geological Congress, held in 2016. Camila Ferreira da Silva and Janderson Bragança Ribeiro review Sobre o Autoritarismo Brasileiro ("About Brazilian Authoritarianism"), by Lilia Moritz Schwarcz. Fabio Mascaro Querido addresses Seja como For: Entrevistas, Retratos e Documentos ("Whichever Way: Interviews, Portraits, and Documents"), by Roberto Schwarz. Mariana Holms analyzes O Homem que Aprendeu o Brasil ("The Man Who Learned Brazil"), by Ana Cecília Impellizieri Martins (2020). Cecilia Marks writes about Franco Moretti's O Romance de Formação ("The Romance of Formation").