It's difficult for the Global South to regulate technologies which are not based in companies, that are based in Brazil. So, if is the United States where the technology is coming from, it may be difficult for Brazilians to protect themselves against the harm that can come from the use of technology. This is the problem we've been seeing all over the world.
. . What is the role of.
. . the regulation of technology outside the U.
S. and China, given that the technology is being developed in the U. S.
and in China. So those are the concerns I think we have to have. It's fundamental, looking from the perspective of a country such as Brazil, which has its own particularities, its own cultural, social, and demographic outlines, it's necessary that we look at these technologies from this perspective, of what we've been calling Global South.
Right? The appropriations of this technology, its development, are still deeply unequal when you look at the capacity of Northern countries, mainly the United States, and the Southern countries. Right now, we are living a unique moment of regulatory turmoil, and also of what we want and intend to have regarding artificial intelligence systems.
Today, the discussion is not only in Brazil, but also in several national, regional, and global forums. So we recently saw how the Organisation for Economic Co-operation and Development (OECD) updated their guidelines on artificial intelligence. UNESCO, which is also a very important forum for a language of human rights, and fundamental for the ethical guidelines for a responsible use of artificial intelligence.
We even saw how some countries, that many times are more reticent in terms of regulation, such as the United States, also through the Executive Order, which acts like a provisional measure, are moving forward in the regulation of artificial intelligence systems. We haven't found a way to effectively regulate, to prevent, and reduce the amount of harm that is caused by the use of technology. Of course, we love our technology.
It's very helpful for education, for public resources, for entertainment, and healthcare, and so forth. But we also have to be concerned about the use of data, specially without consent, in ways that can be harmful to those about whom the data is collected. And is that problem of protecting the data's subject that is so central to our work in the United States, but also so central to what is happening here in Brazil.
The focus on AI regulation is very important, and I think within this is very important to figure out how to address the issues related to the impact of AI on fundamental rights. Because while for safety security we have some expertise, as regard to fundamental rights we have not at the moment, some specific guidance in human rights impact assessment. So for this reason, human rights impact assessments are the core of many ongoing proposals in Europe, in Brazil, and also in other countries that are discussing the future regulation on AI.
Of course, is very challenging not only to set the general framework of AI regulation, but to set the framework for impact assessment. For this reason, it's important to introduce a set of specific key parameters. In the Brazilian proposal now there are specific parameters about the sovereignty, and about the likelihood of the impact, which are two elements that we should consider when we talk about AI.
In the AI proposal, in the European context, the AI Act, so called, is a bit more complicated, it's not the same level of clarity that we find, for instance, in the Brazilian draft for the moment. But also in Europe, there is a huge debate on how to figure out this impact assessment, and the core principle. Of course, after the adoption of the regulation there, we need to implement this principle, provide models, and this could be a task also for the supervisory authorities, in order to support companies and the public sector, in order to put in practice these criteria, and carry proper impact assessment.
A particularity, a Brazilian flavor, is very important in this regard. Differently from other proposals, we can see that the Brazilian proposal aims to be more affirmative of rights. And this unfolds since the way it's structured, so first, it announces the rights, to then announce the duties of the artificial intelligence developers.
It's a proposal that understands we live in a country that is filled with structural asymmetries, such as race and gender, and many others of socioeconomic cuts, it intends to enunciate not only the principle of direct discrimination, but also the indirect. This is a very important issue on the discussions we've been having, because the developers of artificial intelligence systems don't aim to intentionally discriminate those people, but most times, these biases are indirect because of the "wrong" data these artificial intelligence systems learn to automate decisions. So, from matters such as having selection processes that reinforce gender discrimination, mostly of women and black people, to the public security sector, as we have seen through facial recognition, which approaches mostly black people, with several cases happening in Brazil as a whole.
Up until about a year and a half ago, the conversation in my own country about AI was about AI in general, and the ways in which AI might facilitate things like facial recognition technology being used that doesn't work so well on darker skinned people, that could lead to horrible issues around social justice for black people. A whole bibliography and the role of Brazilian civil society that demonstrate how facial recognition fails most precisely with black people, Latin American people, populations that include children and teenagers, because they are growing up, and this will directly affect the accuracy of these technologies. And that also fails the LGBTQIA+ people, especially trans people.
So facing this scenario, the attempt to solve a problem creates new layers of inequality, and new layers of discrimination against these already vulnerable populations. But there is also a big expectation from the society regarding the benefits this technology can bring. Whether to optimize production processes, or develop problem-solving mechanisms for real issues that communities face.
Therefore, we're dealing with a technology that has many capabilities, but thats also carries many risks involving its use, mostly from the perspective of deepening certain inequalities due to the unequal appropriation of this technology. But there are other problems as well, with biases being built into the technology. I did a little experiment where I decided to give.
. . to ask the question of whether I should name my new child.
. . I gave the AI three different names, I gave one that was an African-American sounding name, one that was my own name, and then one that was an European sounding name.
And it described the African-American sounding name as just "unique". And then it described my name, which is Anita Lafrance, as "lovely". And then it described the third name, which was Janet Louise, and European sounding name, it said that name was "classic", and "beautiful".
So it was clearly sort of. . .
It had a stereotype in mind, a bias in mind, to prefer European sounding names to names that sounded like made up black names. So we have. .
. These are just examples. We have ways to go before we can rely upon AI, and drive AI to be bias free.
And rely upon it to provide a resource within reliable use, to fight the causes that we believe in, the social justice causes that we believe in. Today we can see that the digital rights field is mostly white, it has a greater participation and protagonism of the business sectors, including also the government sectors. But the participation of people who are in the basis, of the subjects who suffer the negative effects of the harmful use of their personal data is much more limited.
The discussions about race and gender are happening, but still in very specific spaces. It's still possible to notice big commissions, big important spaces for decision making on digital rights regarding data protection, that are mostly formed by white subjects, and in many cases, exclusively male. This is a matter that needs changing, as fast as it can.
That's why the participation of society is so important. Not only in the discussions about the principles that need to be observed for the use of this technology. Principles guided by ethics, by the possibility of these systems being explainable to the society, so that people can understand their relation with the technology, which apparently is neutral, but that can carry a number of biases.
So there is a very deep and multifaceted discussion regarding these aspects, which have an impact on the regulation discussion. That's why is so important for society to get involved. We wanted to convey the message that it is crucial to involve civil society in the AI agenda, at the national, regional, and international level.
And when we talk about civil society, we should clearly point to those parts of civil society that are usually underrepresented. So it's not enough to have academia in the debate. It's not enough to have digital tech organizations.
We need to give a voice to underrepresented groups, like minorities, vulnerable groups, those that we believe, usually. . .
They may not have something to say, but because they are the rights holders, even if they are not tech experts, they give their essential contribution to the debate. Just as any disruptive technology, the artificial intelligence has a number of issues involving a perspective that we can classify as interdisciplinary, but not only thinking about the scientific and technological development, but also the multisectoral one, because the artificial intelligence has an impact on everyone's lives. But, as I said before, precisely because it is a disruptive and frontier technology, the discussions about its development, its implementation guidelines and regulation, involve more technical sectors and specialized niches.
And I believe it's essential for us to broaden this discussion. To communities, to traditional social movement organizations, to other knowledge fields that are not directly related to the artificial intelligence development. To talk about the creation of technologies, and their impact in the Brazilian context, necessarily involves thinking about the inequalities within Brazil's structure.
Many times, in this innovation framework and attempts to solve problems, we can easily slip into a structure where we try to solve these problems through technology, without dealing directly with the roots of these issues. The challenge of thinking about the connection between misinformation and artificial intelligence happens because the artificial intelligence multiplies and amplifies the possibilities of disseminating misinformation, just as it amplifies the capacity of editing and disseminate truthful and reliable content. But perhaps the biggest challenge for us is how hard it gets to be able to identify authorship, identify liability, allow for people as citizens to understand if that content they're consuming was made by artificial intelligence, whenever this comprehension is key to interpreting the message.
I'll tell you a funny story. I was interested in testing out ChatGPT, and I was asking my little "robot" about legal cases involving African Americans, and I said "Can you please give me an example of five legal cases involving African Americans and their privacies? ", so it immediately gave me five cases.
Then I asked for five more cases, and it gave me five more cases. And I asked for five more cases, it gave me five more cases. I had fifteen cases, I was amazed.
Fifteen cases about black people, and privacy in the United States. Well, in the next day, as a lawyer, I went to research these cases, and they were all fabrications, it was all just garbage. None of them were actually real cases.
So I thought one implication is that generative AI can impede the ability of us to do the research we need to do to effect change. Ironically, the technology may slow us down, give us false information, take us down to rabbit holes that make no sense. So it's not ready for prime time yet, when it comes to certain kinds of social movements, and social progress that we might get through research.
To advance in these discussions, and under a critical view of the technologies, doing so in a perspective that takes into account how the Brazilian society was built, and how it is established up to the present, having as basis structural inequalities of race, gender, and social classes, requires a joint effort, among different sectors of society. It's not possible to think that the public sector is entirely responsible, or even that any problem could be solved through new legislations. We already have legislations, and we need to implement them, at the same time we need to recognize that, thinking about today's technologies and how we comprehend them, while also thinking about the fundamental rights of Brazilian society, are not matters that are misaligned.
They align, and need to be considered as a whole so that we can move towards a future where these technologies can be aligned with the quest for social justice and a more equal society. And this is the way we should follow to ensure that all the care and caution are being taken, so that AI can help us strengthen the informational environment, strengthen the integrity of information, reliable information, but at the same time, that we don't have a misinformation pandemic being spread out. And we need to be prepared to address this discussion aiming to create a legislation that manages to, simultaneously, boost the country's economy, but that also has safeguards to mitigate negative effects, negative externalities, and the risks that come with these technologies.
I think it's a timely topic because data and AI are in Brazil, and also at global level, very important. In Brazil, the recent adoption of the Data Protection Law, the implementation through the Data Protection Authority, and the AI Act under discussion are two key elements in the future scenario of our digital society. But not only in the regulation discussion, it's important for society to also get involved in discussions about the public policies that must be developed with the use of artificial intelligence applications.
Which guidelines the Brazilian government must adopt so that we are no longer just a consumer of this type of technology, but also become a producer of technologies, or a producer of artificial intelligence systems that develops a production chain of the artificial intelligence that must start in many steps of this production process. Starting with the computational capacity, to be able to run robust AI models, until the very own development of AI applications, the monitoring of these applications on concrete reality. So there are many aspects, and therefore, the society's involvement is very important.
