Skip to content
[X]

Analysis

Ximena Mejía Gutiérrez

The position of Latin America in the debate on the regulation of Artificial Intelligence (AI): Who is setting the rules?

- Latin America is being left on the sidelines in the ethical debate around the use of AI in productive activities. What consequences will this delay bring?

The position of Latin America in the debate on the regulation of Artificial Intelligence (AI): Who is setting the rules?

The ethical debate surrounding the development of artificial intelligence (AI) is not new and its origin dates back to the 1980s in the United States. Between 1982 and 1990, a total of $400 million dollars was invested in order to revolutionize the dynamics of "deep learning" [1] and the imitation of the human decision-making process by automated systems [2\ ]. This motivated the emergence of a series of moral and ethical questions about the impact of said technology in the daily life of human beings and the height of such debate can be placed in the year 2016 [3]. It is at this time that States, private entities and civil society organizations begin to produce a series of publications on their position on the design and implementation of AI, with the aim of establishing the main points that must be regulated to avoid human rights violations.

However, not all regions of the world have had the opportunity to contribute, to the same extent, to establish their main concerns about the application of such technology. As Latin Americans and human beings vulnerable to the impacts of AI, it is necessary to ask what role Latin America (LAC) plays in defining the ethical debate on the regulation of artificial intelligence today [4].

Although countries such as Mexico, Brazil, Argentina, Colombia, Chile, Colombia, Ecuador, Costa Rica, Paraguay, Peru, the Dominican Republic and Trinidad and Tobago [5] have shown significant progress in this matter, at the regional level there is no position in which the ethical principles that should govern the use of AI are established. Therefore, it is of vital importance that the States of LAC assume the joint commitment to increase their participation in this controversy, with the objective that the binding documents that are developed in the future incorporate the Latin American perspective. For this, it is suggested to include in their demands postulates of the critical decolonial approach, since this is the main theoretical contribution from said territory. Likewise, the Latin American States are required to coordinate to prioritize the protection of the human rights of the most vulnerable groups in the face of personal data processing processes, and address the future of employment in the region in light of the rise of automation processes. .

In this sense, the work of Jobin, Marcello and Vayena allows us to confirm that the countries of the Global North lead the way in establishing the main ethical problems surrounding the application of AI. The researchers conducted a review of the existing soft law corpus [6] on AI ethics and, in terms of geographic distribution, the results show a prominent representation from the most economically developed countries. Taken together, the United States (25.5%) and the United Kingdom (15.5%) account for more than a third of all ethical discourse produced on the subject, followed by Japan (4.8%), Germany, France, and Finland (each, represents 3.6%). In turn, the European Union also shows an important participation in the delimitation of the ethical discourse of AI, since in this period it generated 7.1% of soft law mechanisms, while some of its member states published 8.3%. For their part, Latin American and African countries and several Asian states are not represented independently of intergovernmental organizations such as the UN [7].

This is not only problematic due to the neglect of the principles of cultural pluralism and the demands for global equity, but also because whoever defines the terms of the debate has the power to establish the priority problems that international laws will address to regulate this technology. which do not necessarily correspond to LAC priorities. The aforementioned authors identified the existence of 11 ethical principles among which the promotion of transparency, non-maleficence, responsibility and privacy predominate. In general, the application of such postulates emphasizes the preservation of privacy, dignity, autonomy and individual freedom; that is to say, the characterization of the debate occurs around the protection of the individual against the design and application of this technology. These principles are of the utmost importance for people to recognize their rights against private entities that, for example, use their information to generate economic income through personal data processing processes. However, where is the responsibility of the most developed States and large private entities manifested towards the commitment to contribute to the economic development of the most vulnerable countries and their populations? Where is the need to develop inclusive strategies for the prevention of job loss and the unfair distribution of the economic benefits that the application of AI will bring, through the extraction of information from groups historically vulnerable?

The inclusion of the critical decolonial approach in this controversy makes it possible to put on the table the responsibility of the most developed States and the largest private organizations regarding the negative impacts of the application of artificial intelligence in Latin America—this in order to promote the generation of inclusive strategies that mitigate such consequences on the Latin American population. Thus, Quijano's notion of "coloniality of power" allows us to understand the continuity of power dynamics between the most developed countries and those that have been called "developing"—through historical processes of dispossession, slavery, appropriation and extraction of resources, which were central to the rise of the modern world— [8]. Therefore, coloniality is what survives colonialism and, for Maldonado Torres, this implies the reproduction of economic, geopolitical, race and gender hierarchies, which were generated as tools of colonial control [9]

From this perspective, the implementation of personal data processing processes supposes a new form of coloniality of power, since the organizations in charge of carrying out these processes convert any aspect of people's lives into information and, at the same time, time, it becomes merchandise [10]. What is concerning here is that, just as in the 16th, 17th, 18th and 19th centuries where colonial powers categorized and excluded individuals based on factors such as racial and ethnic affiliation, AI algorithms risk perpetuating these dynamics in the region [11]. The algorithms developed to automate processes and trained under the values of a society that was disrupted by the dynamics of racial segregation and exclusion tend to replicate racist and discriminatory biases [12]. It is known that LAC is home to a population characterized by its multiethnic and multicultural diversity, so it is not trivial to emphasize the need for the region to insert itself into the ethical debate on AI to prioritize the protection of its population against said colonial biases. .

On the other hand, the Inter-American Development Bank (IDB) warned in 2018 that between 36% and 43% of jobs could be lost due to artificial intelligence as a result of automation [13]. This contributes to increasing the socioeconomic vulnerability of the region and, despite the fact that organizations such as the Massachusetts Institute of Technology (MIT) argue that the loss of vacancies will be compensated by the creation of new jobs [14], it must be taken into account Keep in mind that these will require high degrees of specialization to cover the new needs of the labor market. Therefore, if the necessary questions are not asked to motivate the generation of strategies that mitigate these consequences, we will face a scenario where the concentration of wealth in a few hands will continue to be constant, as well as a context of rampant unemployment.

As Lee mentions, part of the answer will involve training people in tasks that involve creativity, planning, and multidisciplinary thinking [15]—these are skills where AI has proven not very good. However, it is necessary to question who will be able to pay for training in the development of these skills and if there will be a sufficient number of vacancies to absorb the entire job offer.

Therefore, it is necessary for LAC to introduce into the debate on the ethics of AI the questioning of international cooperation mechanisms and international policies, which are required for the use of said technologies to obey a sustainable logic. Assuming that their participation is taken into account, another relevant point that must be put on the table is the democratization of resources and skills so that their population can adapt to the new requirements of the labor market.

Also, it is of vital importance that Latin American leaders contribute to redefine the role of the individual in their interaction with artificial intelligence tools. Within the framework of the forum "International Conference of Data Protection and Privacy Authorities" emphasis is placed on the importance of informed consent when people provide personal data to be processed by AI systems [16]. However, this solution is based on a vision of individual freedom, which does not take into account the inequality of knowledge, resources and, in general, power between AI designers and the rest of the population. Taking into account that the operation of these systems is highly complex—even for their own developers—, it is unfair that individuals are given the responsibility to understand the risks of using such technologies while respecting their human rights.

Sources

    [1] El llamado “aprendizaje profundo” o deep learning es un algoritmo que imita el aprendizaje humano. Este no requiere reglas programadas, sino que mediante una fase previa de entrenamiento, el sistema es capaz de aprender por sí mismo. Smart Panel (2020). ¿Qué es el Deep Learning?. En Smart Panel. Recuperado de: https://www.smartpanel.com/que-es-deep-learning/

    [2] IA LATAM. (s.f.). Historia de la Inteligencia Artificial. En IA LATAM. Recuperado de: https://ia-latam.com/2019/02/07/historia-de-la-inteligencia-artificial/#:~:text=En%20la%20d%C3%A9cada%20de%201980,computadoras%20aprender%20usando%20la%20experiencia.

    [3] Jobin, A., Ienca, M., & Vayena, E. “The global landscape of AI ethics guidelines. Nature Machine Intelligence. Vol. 1 (2019), https://www.nature.com/articles/s42256-019-0088-2

    [5] Banco Interamericano de Desarrollo (BID). La Inteligencia Artificial al servicio del bien social en América Latina y el Caribe. (Washington: BID, 2020, 16).

    [6] En contraste a la llamada hard law (es decir, regulaciones legalmente vinculantes), la soft law no es vinculatoria, sino que tiene un carácter persuasivo y contribuye a delimitar instrumentos legales vinculantes.

    [7] Jobin, A., Ienca, M., & Vayena, E. “The global landscape of AI ethics guidelines. Nature Machine Intelligence. Vol. 1 (2019), https://www.nature.com/articles/s42256-019-0088-2, 3.

    [8] Quijano, A. “Coloniality of power and Eurocentrism in Latin Americas”. International Sociology, 15, núm. (2000). https://journals.sagepub.com/doi/10.1177/0268580900015002005

    [9] Maldonado-Torres, N. On the coloniality of being: contributions to the development of a concept. Cultural studies, 21 núm 2-3, (2007). https://www.tandfonline.com/doi/abs/10.1080/09502380601162548?journalCode=rcus20

    [10] Couldry, N, Mejías, U.The costs of connection: how data is colonizing human life and appropriating it for capitalism. (California: Stanford University Press, 2020), 109

    [11] Los programas de “machine learning” tienen como objetivo detectar patrones en los datos que se les proveen, para luego replicarlos en predicciones futuras. Es por ello que la información que devuelven reflejará los sesgos presenten en la información que se utilice para alimentar el programa; esto propicia que la tecnología se convierta en un legitimador de viejos estereotipos y dinámicas de discriminación. Scrollini, F. Automatizar con cautela. Datos e Inteligencia artificial en América Latina. (Costa Rica: ILDA, 2018), 1.

    [12] A manera de ejemplo, puede citarse el caso del programa “PredPol” en Uruguay, cuyo fin era generar mapas predictivos sobre dónde ocurririrían incidencias delictivas en un plazo de 24 horas. La base sobre la cual “PredPol” realiza sus predicciones es el registro de datos de criminalidad del Ministerio del Interior. El software dejó de utilizarse, debido a que un informe publicado por la organización Human Rights Data Analysis Group sugirió que este tipo de programas genera un «circuito de retroalimentación». Esto lleva a los policías a las mismas zonas de la ciudad, generalmente en donde se concentra mayor cantidad de minorías raciales, independientemente de la verdadera tasa de criminalidad en esa área. La lógica del “circuito de retroalimentación” fue alimentada por los sesgos discriminatorios del viejo olfato policial. Scrollini, F. Automatizar con cautela. Datos e Inteligencia artificial en América Latina. (Costa Rica: ILDA, 2018), 13.

    [13] Schechter, P. Latin America’s Growing Artificial Intelligence Wave. En Brink The Edge of Risk. Recuperado de: https://www.brinknews.com/latin-americas-growing-artificial-intelligence-wave/#:~:text=Apocalyptic%20scenarios%20aside%2C%20the%20immediate,as%20a%20result%20of%20automatization.

    [14] Massachusetts Institute of Technology. The Global IA Agenda: Latin America. (Massachusetts: MIT), 4.

    [15] Lee, K. The Real Threat of Artificial Intelligence. En The New York Times. Recuperado de: https://www.nytimes.com/2017/06/24/opinion/sunday/artificial-intelligence-economic-inequality.html

    [16] IPANDETEC. Public consultation ethics and data protection in artificial intelligence: continuing the debate. a contribution from Latin America & the Caribbean. En IPANDETEC. Recuperado de: https://www.ipandetec.org/2019/04/12/public-consultation-ethics-data-protection-artificial-intelligence/


The best content in your inbox

Join our newsletter with the best of CEMERI

Related articles

Mejía, Ximena. “La posición de América Latina en el debate sobre la regulación de la Inteligencia Artificial (IA): ¿Quiénes están poniendo las reglas?.” CEMERI, 9 sept. 2022, https://cemeri.org/en/art/a-inteligencia-artificial-america-latina-bu.