By akademiotoelektronik, 21/09/2022

Artificial intelligence must become everyone's business

Mieke de Ketelaere fell into the pot of new technologies when she was 13 years old.""My first computer was a Texas Instrument Ti-99/4A, says this 50-year-old Gantoise domiciled in Walhain.I could quibble there for hours in my room.I understood quickly that, thanks to technology, it was possible to automate certain tasks to free up time and improve well-being.""

After studying industrial engineering, she continues with a master's degree in robotics and electromechanics in Stuttgart, before starting a doctorate in artificial intelligence in Auckland, New Zealand.The first part of his professional career takes place in the computer sector and data analysis (Microsoft, SAS, etc.).

In May 2019, she joined the IMEC, Flemish Research Center applied in nanotechnologies, to direct artificial intelligence programs.Since the beginning of May, she has been Director of the AI strategy in Imec-Idlab (based in Ghent and Antwerp).

Votre livre est un plaidoyer en faveur de ""traducteurs d’IA"".Who are these artificial intelligence translators (AI)?

To answer you, you have to explain the reason that pushed me to write this book.Originally, my project was to write a book on the role of a woman in the business world and on the way this world can take on women.The publisher found it interesting, but he wanted me to write a book on AI first, which is my field of competence.What I refused.But it is that, shortly after, I found myself in a debate on robotics in which Alexander de Croo participated. Interrogé sur la question de la responsabilité lorsque l’IA est à l’origine d’une erreur, il a répondu : ""C’est la responsabilitéde l’ingénieur.""I didn't agree!

Why ?

Because the engineer who develops artificial intelligence does not always know where and how his algorithm will be used.This story made me think about the lack of understanding of AI.I went back to the publisher to say that I was ready.My idea was as follows: as an engineer, we have to look at ourselves in the mirror and take responsibility.With AI, we have advanced technology that progresses very quickly, which spreads everywhere, but that unique people do not understand.In the world of business, it goes a bit in all directions.Business leaders believe that by hiring a data scientist and collecting as much data as possible they will be able to do great things with AI.This is not how things are going!The idea of the book is therefore part of there: how to find a common language that can help people speak and understand each other when we talk about AI.For that, we need AI translators.

Who are they ?

It's everyone!Everyone, with their knowledge, their expertise, can become it.People from the world of business, research, philosophy, law, ethics, civil society ... Everyone must be able to express themselves on what they expect or not from theIa, what he wants to do with it.You have to get AI from the world of experts, technologists, so that the population appropriates it.AI must become everyone's business.It is not a question of becoming all experts in AI, but of understanding the general principles and the consequences.Take the example of the car.Everyone knows that you have to have a driving license and you have to attach their belt to avoid serious consequences.We became aware of the dangers of driving a car.With AI, which we find today in our smartphones or on Google, most people do not know much.The difficulty is that AI is invisible technology.

Parmi les personnes qui connaissent l’IA, on observe un clivage entre ""adorateurs"" et ""alarmistes"".

You have to get out of this cleavage.A median route exists.Today, we have a lot of examples that show that AI can have a real added value.It's not about stopping everything.But we also see that we have come close, even sometimes crossed, societal or ethical limits.There is the question of biases (gender, race ...) which alter algorithms and that of overconsumption of energy.To resume the analogy of the car, we did not stop using it in the 1950s because we witnessed an increase in accidents.We rather looked for a framework to limit these accidents (highway code, security of vehicles, etc.).AI is the same, except that it advances very quickly and that it is invisible.

The European Commission tabled, at the end of April, a proposal for regulation on AI whose objective is to supervise its use in various areas of application.How do you judge this proposal?

Unfortunately, the Commission wrote a proposal valid for the 1980s.It is a matter of the principle that AI was embodied in clearly identifiable algorithms and uses.However, AI, data and algorithms are a whole chain, with multiple interactions.We are not in a situation where an AI evolves independently, that is to say independently of its environment.So, I fear that this European proposal will be in total lag with the current situation and that, moreover, it hinders the innovation permitted by AI.Europe has forgotten to look at what was being done in terms of AI and what it was going to become in five or ten years.On this one, it might have had to consult the engineers in AI.

When your friends challenge you on AI and its potential drifts, how do you reassure them?

Really worried people are often after seeing or read what is said of AI in the media.This is the Terminator version of the AI!I explain to them that it is enough to remove the battery or the plug to stop everything.It is a technology created by humans and on which man keeps control.But, in 90 % of cases, people are not particularly worried.In fact, they just don't know what AI is.This is where I say that you absolutely have to be interested in it.The media can play an important role by giving an explanation and a more balanced point of view of the AI.This is the case, for example, about the impact of AI on employment.We have already known, in the past, technological innovations and they have had an impact on the nature of the jobs.But there is never a pure and simple destruction of employment.What we have known are job transformations.We are going to experience the same transformation with digital and AI.Covid's pandemic is a very good example.Some transformations have accelerated.But we also saw that robots or machines, even intelligent, were perfectly incapable of taking blood, assisting oxygen patients, accompanying them in their revalidation ... AI had nothing predicted and itwas not very useful to cope with the pandemic.This shows all the limits of AI and the need to demystify it.

AI is not very intelligent ...?

It is a weak AI, as the experts say.She has the intelligence of a 3 or 5 year old child.She is able to learn, to calculate, to repeat very quickly.But she is unable to reason, to collaborate, to have empathy or even simply make the difference between a real snowman and a child disguised as a snowman.

The main challenge of AI, write, is not really technological.He is mainly responsible.What is a responsible AI?

It is not because a technology works automatically to use it and deploy it on a large scale.We must take into account the context.Should the algorithms used in China, for example, be with us without questioning their impact?We know that no.We need to counterbalance the discourse of technologists and engineers by other points of view.It's really urgent.Everyone, including engineers, should look at themselves in the mirror and take responsibility.You have to dare to ask for explanations when you don't understand.We must absolutely understand each other by speaking the same language.For that, you have to create bridges between people from different universes and stop working in silos, to make an inter-self.

His book

Man versus machine.Artificial Demystified Intelligence, Pelckmans, 240 pages, May 2021, 30 euros.The book was published in Dutch in November 2020 and will soon be available in English.Mieke de Ketelaere decided to give up her copyright to the DWENGO project.children's awareness org of technologies.