The Unesco member states now have a guide to “deal responsibly with the known and unknown effects of artificial intelligence on human beings, societies and the environment and ecosystems”. That is the goal of the Recommendation on the Ethics of Artificial Intelligence, a sort of universal statement for the use of these systems that during the general conference held this Wednesday was adopted by the 193 countries of the organization. The guide considers the principles that should inspire the development of such controversial technologies as facial recognition and also addresses less controversial topics, such as the potential effect of recommendation systems on cultural or linguistic diversity. “This text establishes for the first time a global regulatory framework for the use of artificial intelligence. It is based on three pillars: respect for human rights, the rule of law and the fight against discrimination ”, explains to EL PAÍS Audrey Azoulay, general director of Unesco.
The 30-page document that Azoulay describes as a “historical text” is not legally binding, but Unesco hopes that it will become a global reference for the development and ethical use of this technology. In this context there are already numerous frameworks and guides developed by organizations at different levels, from the initiative of entities such as the OECD to proposals such as the digital rights charter adopted by the Spanish government on July 14. “Unesco asks its 193 member states to take all necessary measures to apply this ethical framework. The Organization will periodically evaluate its implementation, asking States to report on their progress and practices in this area. These results will be analyzed by experts and submitted to public debate in a transparent way ”, Azoulay points out.
Expert groups made up of representatives from Google, Facebook, Microsoft, Stanford and New York Universities, the Chinese Academy of Science and Technology, have participated in the drafting of the document, which began in early 2020. In September last year, a first draft was presented and the deadline was opened for Unesco countries to submit their comments and observations on the text, which has continued to be revised until the last general conference.
Among the keys to the recently adopted recommendation is the definition of basic principles that concern the development of some of the most questioned technologies of the moment, such as facial recognition systems. According to the document adopted by UNESCO, which has countries like China among its members, these systems “should not be used for the purposes of mass surveillance or social accountability.” In the case of autonomous weapons, the Unesco Deputy Director General for Social and Human Sciences, Gabriela Ramos, explains in a statement that, although there is no explicit reference to drones for military use or lethal autonomous weapons, the recommendation establishes basic rules such as that “life and death decisions should not be made by artificial intelligence systems” and it is specified that in these scenarios “the last word must be human”.
The recommendation also addresses less extreme but equally relevant scenarios, such as the potential effect of the recommendation systems used by the platforms of streaming like Netflix or Spotify in cultural and linguistic diversity. A recent report from the European Union pointed out that the criteria used by these models to offer new content “are not transparent or auditable” and probably base their decisions on economic factors that benefit the platforms. “Local artists or underrepresented creative and cultural content are less likely to appear in the recommendations of these systems if their work or shows are insufficiently lucrative,” the report notes. In this context, the recently released guide calls on countries to work on a better understanding and evaluation of both the positive and potentially harmful effects of recommendation systems.
With a view to a hypothetical but increasingly plausible future where people interact with robots and artificial intelligence systems capable of recognizing and reproducing human emotions, the document requires that the ethical implications of these relationships be evaluated, especially when they are established with minors. . “The constant interaction with artificial intelligence technology, including social algorithms, could subject children and adults to manipulations and negative consequences for their mental health,” continues Ramos. On the other hand, the need for individuals to be informed at all times whether they are dealing with one of these systems or are subject to decisions made by them is underlined. It is also requested that mechanisms be established to access the reasons that guide a decision that affects the rights or freedoms of users, as well as ways to challenge it or request a reevaluation.
Other matters included in the framework of this recommendation are the protection of privacy, the prevention of biases that may harm the performance of these systems, the promotion of a more inclusive environment in the sector responsible for developing these technologies or even the reduction of the environmental impact of these systems. According to a study published in the journal Nature, training an artificial intelligence algorithm can produce about 300,000 kilos of carbon dioxide, which is equivalent to 125 round-trip flights between New York and Beijing.
In addition, the document that has just been adopted by UNESCO requires governments to educate their citizens in the digital rights it seeks to protect. “By educating a new generation of digital consumers aware of their rights and with the critical thinking skills necessary to navigate the digital sphere, we will be empowering them to hold big tech to account and we will be demanding that AI be developed in a way. human and ethical ”, Ramos sentence.
You can follow EL PAÍS TECNOLOGÍA at Facebook and Twitter or sign up here to receive our newsletter semanal.