How can artificial intelligence discriminate?

AI helps determine what treatments will be used on sick people, determines who will qualify for life insurance, diagnoses tumors more effectively than humans, indicates how long in prison a person will have to serve, or which applicants in different jobs should get one. Interview. From simple and often irrelevant applications, it was implemented in places that could become dangerous if the necessary precautions were not taken.

Although it happens relatively frequently, the latest case of discrimination in the world of artificial intelligence involved Amazon, where an algorithm in charge of selecting the best talents had learned to penalize resumes that had the word “woman.” But it was not the only one. In 2015, Google’s facial recognition technology identified certain African-American people as gorillas, or even a Microsoft chatbot, which began tweeting racial slurs after learning from the things they said, imagine these skills updated in 2021.

Such is the concern about these episodes that some of the main leaders in the discipline are working to reverse this process. Fei-Fei Li, who worked as chief AI scientist at Google, was the director of the Stanford AI Lab, created AI4ALL, a non-profit organization that intends to increase diversity and inclusion in the world of artificial intelligence, revolutionized The technological world with the creation of ImageNet, and a practically endless CV, is one of the people who are trying to change the path that, in recent times, the development of this technology has taken.

While some charlatans are talking about how artificial intelligence will conquer the world, she is focused on things much less melodramatic, but more real and current: on how this technology will affect people’s work and lives and how it will modify the human experience. For it to work, fundamental changes must be made from development. If we succeed, advances in this field will be a transformative force for good. If not, they will leave much of humanity out of the equation.

In a speech to the US Committee on Science, Space, and Technology, she said:

There is nothing artificial about artificial intelligence. It’s inspired by people, it’s created by people, and most importantly, it affects people. It is a powerful tool that we are just beginning to understand, and we have a deep responsibility.

Li realized the inequality in the field of artificial intelligence between men and women at an awards ceremony. And despite what the majority think about this type of technology, that being pure mathematics and training they are practically not wrong, it is not like that. These algorithms follow a series of rules, and those rules are chosen by human beings. Without a diverse group of engineers, Li explains that we will end up with biased algorithms making unfair decisions on, for example, who they could give a loan to, or neural networks trained only with white faces that will end up malfunctioning in non-white people — It applies to Latinos, Asians, African Americans, etc. — as happened with Google.

Discrimination by machine learning and artificial intelligence algorithms will not go away if you don’t work proactively to make that happen. Companies have to be in charge of creating diverse groups not only of engineers but of various disciplines, ethnic origins and genders in order to avoid these types of problems from being reflected in technological developments. If this does not happen, the great problem for the future of artificial intelligence will not be that it conquers the world in the Hollywood style, but that we will have created a technology developed to generate exclusion with potentially irreversible consequences.

What you have to do, as with any new development, is to put people first and understand that, although it seems impartial, this technology is not because the people who develop it are not. You have to think about that and work so that artificial intelligence shapes the world we want it to be, beyond how the world is today.

Cathy O’Neil, in her spectacular Weapons of Math Destruction, explains it perfectly:

Big data processes encode the past. They do not invent the future. Doing that requires moral imagination, and that is something only humans can provide. We have to explicitly integrate values ​​into our algorithms by creating models that follow our ethics. Sometimes that is going to mean putting fairness above profit.

This is one of those moments when it is better to take a step back, understand what it takes to make this technology enrich our society and only then begin to innovate, returning to the drawing board as many times as necessary. It is not the way you usually work in technology, but it is a good way not to screw it up and deepen the injustices that affect millions of people.

Related posts:

Lascia andare le persone che non sono pronte ad amarti. Questa è la cosa più difficile che dovrai fare nella tua vita e sarà anche la cosa più importante. Smettila di fare conversazioni difficili con…

Party Games Party games are great to start conversations and let guests get to know each other. There’s a variety of games like “Would You Rather?” that can spark discussions and even debates. Don’t…

El equipo entra en una taberna de la zona pobre de la ciudad. El olor a alcohol es intenso. El tabernero saluda con voz ronca, intentando fingir todo el entusiasmo posible. Limpia una parte de la…