A group of European Union experts this week released a document with ethical guidelines to guide the development of artificial intelligence technologies. These are used in various activities, from the selection of publications shown on social networks to credit rating systems.
The term artificial intelligence is used to characterize systems that offer services based on a sophisticated information processing capacity, similar or even superior to that of human beings. Certain technologies have already performed better than people, whether in a game, in a diagnosis or in the recognition of images or texts.
The document is part of a European Union process to establish a “human-centric” vision for building “reliable” solutions for this type of systems. This can materialize both in legislation and administrative rules and in general guidelines for manufacturers and research projects.
One guideline is the relevance of the participation and control of human beings, with technical objects that promote the role and rights of people, and do not harm them. A complementary guideline is the guarantee that the systems consider the diversity of segments and human representations (including gender, race and ethnicity, sexual orientation and class, among others), avoiding actions that generate discrimination.
According to the document, AI systems must be “robust” and “safe”, in order to avoid errors or be able to deal with them, correcting any inconsistencies. These problems can have serious impacts on society, such as discrimination against people in accessing a service or even stock market crashes whose purchases and sales of shares use these technologies.
At the same time, the text highlights the need to ensure the transparency of systems. This is because the opacity of these technologies can bring risks, since their intelligent character makes it more difficult to understand why a certain operation or decision was taken in a certain way and not in another. Thus, the authors argue that an AI system should be “traceable” and “explainable”, so that there are no difficulties in understanding its performance.
According to the document, these technical solutions must ensure citizens' privacy and control over their data. Information collected about an individual cannot be used to harm that individual, such as in automated decisions that discriminate against someone else. Studies have already shown how these technologies can incorporate biases, privileging, for example, white people over black people in characterizing or offering a service.
The Vice-President for the European Union's Digital Single Market, Andrus Ansip, highlighted the importance of the topic. “The ethical dimension of Artificial Intelligence is not just a luxury or an addition. It is only with confidence that our society can fully benefit from these technologies. Ethical AI is a win-win proposition that could be a competitive advantage for Europe: to be a leader in people-centric AI technologies that users can trust.”
Source: Agência Brasil
The term artificial intelligence is used to characterize systems that offer services based on a sophisticated information processing capacity, similar or even superior to that of human beings. Certain technologies have already performed better than people, whether in a game, in a diagnosis or in the recognition of images or texts.
The document is part of a European Union process to establish a “human-centric” vision for building “reliable” solutions for this type of systems. This can materialize both in legislation and administrative rules and in general guidelines for manufacturers and research projects.
One guideline is the relevance of the participation and control of human beings, with technical objects that promote the role and rights of people, and do not harm them. A complementary guideline is the guarantee that the systems consider the diversity of segments and human representations (including gender, race and ethnicity, sexual orientation and class, among others), avoiding actions that generate discrimination.
According to the document, AI systems must be “robust” and “safe”, in order to avoid errors or be able to deal with them, correcting any inconsistencies. These problems can have serious impacts on society, such as discrimination against people in accessing a service or even stock market crashes whose purchases and sales of shares use these technologies.
At the same time, the text highlights the need to ensure the transparency of systems. This is because the opacity of these technologies can bring risks, since their intelligent character makes it more difficult to understand why a certain operation or decision was taken in a certain way and not in another. Thus, the authors argue that an AI system should be “traceable” and “explainable”, so that there are no difficulties in understanding its performance.
According to the document, these technical solutions must ensure citizens' privacy and control over their data. Information collected about an individual cannot be used to harm that individual, such as in automated decisions that discriminate against someone else. Studies have already shown how these technologies can incorporate biases, privileging, for example, white people over black people in characterizing or offering a service.
The Vice-President for the European Union's Digital Single Market, Andrus Ansip, highlighted the importance of the topic. “The ethical dimension of Artificial Intelligence is not just a luxury or an addition. It is only with confidence that our society can fully benefit from these technologies. Ethical AI is a win-win proposition that could be a competitive advantage for Europe: to be a leader in people-centric AI technologies that users can trust.”
Source: Agência Brasil
EN 
PT 
