Why watch Coded Bias ?
Why watch Coded Bias: what documentary brings important for reflection on an ethical AI.
Coded Bias is, for me, so far, the best documentary to address the social impacts of technologies using artificial intelligence. Of course, these documentaries almost always emphasize the risks and problems caused by artificial intelligence and it is understandable, since, in addition to legitimate criticism of the effects of any technology, it seems that the public also has a predilection for the controversial side of things. Coded Bias presents this side very well, which does not prevent us, however, from pointing out ourselves the obvious benefits and advances that AI has already brought and will still bring us.
I think that's the first important point to have an overview of AI, which co-reads the critical bias that the film brings to us with the recognition of the beneficial potential of this technology. Joy Buolamwini, who along with Deborah Raji developed research on bias in facial recognition systems that takes much of the documentary, implicitly recognizes this beneficial potential of technology in general – which is not an enemy of humanity – but warns of the risks of technological promises not to be made if society neglects its ethical aspects. This seems to be a good way to deal with documentaries critical to certain uses of artificial intelligence: we should not leave with the impression that we are moving towards the"end of the world", but with the warning that care must be taken so that AI can serve the well-being of humanity, which is, after all, the fundamental criterion for the use of a technology to be considered in accordance with ethics. The film itself shows how Joy's review brought results: at a conference in South Africa, she presented the improvements IBM made to its facial recognition system, which resulted in a sensible reduction inthe amount of recognition errors of facial biometrics of black people.
The problem of racial and gender biases in facial recognition that the film presents has much to do with the lack of diversity in the datasets used to train machine learning algorithms. Buolamwini and Raji show in their research that the datasets used to train facial recognition systems have almost exclusively images collected in the United States and some European countries where the population is mostly white. So, it's not surprising that there are more mistakes in recognizing black people. AI will be as good as the quality of the data with which it is fed. Hence the importance of ensuring diversity in the datasets used, as well as in ai product development teams.
Another important aspect that the film shows: "data is target". As the data is a reflection of our history, its use carries the risk that we will be stuck in the past, in a looping that repeatedly reproduces what it already is, but prevents or hinders what it may become. In this context, the social changes that are necessary from time to time become the most difficult to happen. Hence the importance of ethics and critical thinking applied to AI, because they can collaborate so that it does not develop in order to cast social advancement.
The documentary also makes a tour of the world, addressing the use of facial recognition in London for public safety purposes and in China where this technology is used as the main means of payment by the population. Facial recognition makes many fear the appearance of something like a dystopia in the style of big brother's book 1984, on the other hand, it is certain that for public security purposes it is a fairly effective technology. What to do with it is still the subject of intense debates around the world. Some cities in the United States have banned the use of this technology, other than defend its use regulated by law that is able to balance security concern with the guarantee of people's rights. Dealing properly with this technology is not an easy task. It will certainly still require many debates and reflections. In Brazil, the Governor of São Paulo recently vetoed a bill that determined the use of facial recognition on subway and CPTM trains in order to curb the action of criminals in these environments. The Governador justified the veto by stating that the project improperly interferes with the freedom of companies that administer these means of transportation in São Paulo, in addition to violating state laws and the Constitution. It is interesting to note that the use of facial recognition in the area of public safety does not yet have legal regulation in Brazil since the General Data Protection Law (LGPD) does not apply to the processing of data performed for public security purposes.
Coded Bias successfully shows how you cannot separate the technical side from the social side, as Joy Buolamwini says. Many are thexemplos of this as the control of the population in China with the use of technology, the difficulties that social networks bring to the functioning of democracies and the use of algorithms to suggest decisions in criminal matters in the United States.
This diversity of social issues that AI brings us shows how it is a technology of general purpose, which reaches the most diverse aspects of life. That is why the importance of ethical parameters that ensure an AI focused on human well-being, and that results in the establishment of governance and regulation of this crucial technology for the future of humanity.
About the author
Doutor em Filosofia do Direito pela PUC-SP. Mestre em Filosofia do Direito pela USP. Pesquisador e autor de publicações sobre inteligência artificial, big data e proteção de dados.
Event summary by Connor Wright, Partnerships Manager of MAIEI.
International Trends in Artificial Intelligence Strategies
On September 28 and October 5, we will be holding webinars that cover trends in Artificial Intelligence (AI) strategies. In the first event, we will bring together experts who will