Artificial intelligence at the Pariscience festival: “it’s a dangerous mythology” – VousNousIls

From October 11 to 21, the school edition of the Paris science festival, which allows classes to attend free screenings of documentaries. In this context, the film Autopsy of an artificial intelligence was presented to high school students. Meeting with its director and co-author, Jean-Christophe Ribot.

Jean-Christophe Ribot

Can you introduce yourself ?

I am a writer-director of documentary films for television. I mainly deal with scientific topics.

Who is the documentary for?

The film is for everyone. The challenge was to make the subject accessible to everyone and to show what the limits of artificial intelligence are.

How did you get interested in AI?

I’ve always been interested in this subject, and I’ve wanted to devote a film to it for a long time. As a teenager, I was a geek, I really liked programming, and I shared this fantasy of recreating an “artificial mind”. But I came back pretty quickly.

Today, I try to analyze the enthusiasm for artificial intelligence, promoted in particular by the way in which this subject is publicized. We often find, even in certain documentaries which are intended to be critical, the idea that the AI ​​​​is an emerging intelligence, with the same capacities as a human brain: it is a mythology that I find dangerous.

What is the danger?

The first danger is to reduce the human brain to circuits ex nihilo, which would simply do calculations. We see this idea in a whole branch of neuroscience, which readily lends itself to this metaphor.

The other danger is implying that the AI ​​has capabilities beyond what it actually has.

In the film, we give a fairly concrete example: when we are interested in the way in which a word takes on meaning in a human being, we realize that the sensitive, emotional experience, which relates to a body in the flesh and in bone, is specific to human beings. While a computer will understand the word “apple” through pixels or other words it knows, humans will understand it thanks to all their sensations: that of the fruit in the palm of the hand, of the taste on the taste buds… Sensory memory and emotions contribute to this intelligence.

In the film, you also address the ethical limits of artificial intelligence. Is AI a “false” progress for humanity?

This is real progress, in the positive sense of the term “progress.” As long as we understand artificial intelligence as a tool – like we would use a very sophisticated calculator. Where it becomes problematic is when it is used as a prescriber of faith to which we blindly rely. Today, we use it to disentangle the true from the false, without knowing how the verdict is rendered. This is the case in the private sphere as well as in the institutional sphere.

For example, we can cite the algorithms of dating applications, which recommend partners, or the algorithms which determine the risk of recidivism of defendants, in certain states of America. This practice raises a whole host of ethical questions: what data is entered into the machine? How are they analyzed? Who decided on the programming? What purpose ?

The first thing to understand is that deep learning algorithms do nothing more or less than statistics. These are complex data analysis systems, but in principle they are indeed statistics: on the basis of data collected in the past (for example, sociological characteristics) we make predictions about the future. We understand the problem well: for a sociologist, statistics are used to show the state of society, to provide a critical look and possibly solutions. But today, algorithms only serve to recommend something based on the past, and therefore reinforce the current state of affairs – whether it is satisfactory or not.

For example, there is a racist bias in the United States, which means that African Americans are given a higher recidivism rate. By taking data from the past, the algorithm reproduces the same bias.

What moral can we draw from the film?

The idea is that we, institutions, humans, can reclaim our choices. Artificial intelligence should be a decision-making tool, but it has become a prescriber of choice, a device for automating individual or social choices. It is important not to always rely on external power, held by those who make the algorithms (who are largely GAFAM: Google, Apple, etc.) We must face up to our responsibilities, and tell ourselves that our choices are ours.

Is AI a particularly difficult subject to popularize?

It is difficult, yes. The challenge for us was to understand, concretely, how artificial intelligence works – without going through metaphors. And computer scientists are not always very good teachers… Once you have the basic principle, you understand that AI is applicable to many fields that are similar to what humans have most specifically ( language, music composition…) but also that it has limits: it can only create a piece of music from what has already been produced by humans in the past.

The difficulty was therefore to essentialize the technological question, to link it with more general questions. When you get there, it’s pretty satisfying.

The film also appears in the school edition of the Pariscience festival: is this a subject in which young people must be particularly trained today?

This is essential, especially for young people because they are heavy users of artificial intelligence (for example with dating applications). The stakes are enormous for them: they have to understand how it works, in order to be able to decide whether or not to follow their prescriptions. If we want to have power over our choices, we have to ask ourselves how the algorithms we use work, but also why, what are the interests of those who propose them… It’s important politically, and it raises the question of freedom. for individuals.

Movie Trailer

We would love to say thanks to the writer of this article for this remarkable material

Artificial intelligence at the Pariscience festival: “it’s a dangerous mythology” – VousNousIls


Visit our social media profiles as well as other related pageshttps://www.ai-magazine.com/related-pages/