Marc Levy: "ChatGPT is a gifted child, very poorly educated"

Since 2019, Marc Levy has been working on writing the "9"

Marc Levy: "ChatGPT is a gifted child, very poorly educated"

Since 2019, Marc Levy has been working on writing the "9". A trilogy being adapted into a TV series by Costa Gavras. The "9" currently consists of three parts: It happened at night, Le Crépuscule des fauves and Noa. It tells of the entry into resistance of hackers who do not know each other, do not live in the same place, but unite to fight the cyber crooks of the 21st century.

This trilogy of thrillers marks a 180 degree turn in the novelist's work. It is also the result of long observation of the technological change in our society since the digital revolution of the 2000s. Implicitly, and barely concealed, we find, in the "9", the characters inspired by the greatest cases of the last twenty years:

Mogul Rupert Murdoch and his son, Lachlan, in the transformation of Fox News, which resulted in the Dominion lawsuit, for arguing that the 2020 presidential election was rigged, knowing it was wrong. We can also imagine the Koch brothers, famous for their disinformation mechanics, since by buying three hundred local press dailies in the United States from 2010, they helped to plunge the country back into conservatism. The trilogy also tackles the UK Cambridge Analytica scandal. With the appearance, this time, of the technique of "micro-targeting" - the dissemination and immediate deletion of fake news on Facebook.

Marc Levy, who has been living for a long time in New York, has become a discreet but attentive watcher of planetary algorithmic developments that allow opinion to be manipulated and colossal financial profits to be made through data collection. New technologies are the subject of his latest novels. We bet that the chatbot ChatGPT, the language model based on artificial intelligence (AI) developed by OpenAI, will take pride of place in the next one.

Le Point: How do you see the arrival of ChatGPT?

Marc Levy: ChatGPT is a gifted child who is learning. But it is not God who informs him! The AI ​​reasons from the sources of information available on the Internet, which come from, among others, newspapers and television channels. From the moment you let this child learn in an ultra-polarized environment, and where misinformation develops at an increasing rate, the probability that your artificial intelligence will be very poorly educated is enormous. That is what is worrying. An AI connected to Fox News or CNews will recycle information and indoctrinate people tenfold.

In what forms are people going to have access to AI? The Snapchat app, for example, has just integrated AI into its services. Can you imagine the danger to young people? The great perversity of the integration of artificial intelligence in the applications available to everyone is that they will obviously replace search engines and, therefore, the way we do research. And, as whoever does not say a word agrees, the most accessible information, the easiest to read, will become true. For example, my 13-year-old son went to ask Snapchat's AI, "Could you tell me who Marc Levy is?" »

Among other things, that I joined the Socialist Party at a very young age and had been supporting the fight against childhood cancer for a very long time. What is wrong.

Do you know how the AI ​​retrieved these fake biographics?

The AI ​​must have found that I loaned my image for a campaign for a children's cancer charity a few years ago. I posed in the association's t-shirt and the campaign aired, at the time, in the United States. I think the AI ​​looked for this image in American and not French databases. Otherwise, instead of outputting this activity which accounts for about 0.003% of my total activity time, it would have mentioned The Red Cross, which I have been genuinely committed to for a long time.

How did the AI ​​label you a socialist?

I don't know. And what scares me is that it's reading my last three novels, the trilogy of 9... But above all, do you realize the consequences of the use of artificial intelligence in totalitarian countries? ? Today, state services in some countries already systematically analyze all information available online about a person applying for a visa. What future for these practices with the help of an AI?

AI could be doomsday, but is that the only danger to society?

The biggest danger, for me, is that it will cause in the next ten years a cultural, societal and political revolution ten or a hundred times greater than that of the digital revolution which changed the world between 2000 and 2023. L he greed, success and pressure from the hierarchy are so important that we start by making and then we think.

The extraordinary mismanagement of the digital revolution is linked to the absolute lack of culture of the people who have contributed to the development of applications intended for young people, by incorporating tracking, geolocation, facial and digital recognition tools, data feedback huge, etc.

Today, your car, your phone or your clock radio, all the home automation devices you have purchased in your home, feed back information about your lifestyle and your behaviors to the nearest second to those who manufacture them. They resell them – I talk about it in It happened at night – to data merchants who allow the agglomeration of this data to reconstitute your profile.

What users don't know is that all of that isolated information, like what time you get up, taken from any device and your IP address, is passed into an index that identifies you. Your profile becomes extremely complete over time.

At the time of writing It Happened at Night, in 2019-2020, the data collection of an ordinary individual amounted to 23,000 entry points. Which was a $7 billion market! Other algorithms then take over. I believe that the important thing is not the power of AI, but the use we make of it.

A 25 km/h car in town is used to get you around. At 100 km/h, it becomes a killing machine. The debate is not whether AI is good or not. It would be a matter of polarization. Where the AI ​​revolution is going to have a tremendous benefit to society is that it's going to be able to think smarter than humans.

If we take the problem of dealing with traffic in a city like Paris, by integrating the constraints of fluidity and pollution, it is certain that the response of the AI ​​will be smarter than what was planned by the Paris City Hall. , because it will work without ideology. Artificial intelligence will protect us from certain wanderings.

I'll tell you something you rarely hear, because it's taken as an admission of weakness, although I think it's the opposite: I don't know. And I don't think we know. When I was 25 or 30, I was making music with a friend, and we spent all our savings on synthesizers. Suddenly we were able to reproduce the sound of fifty violins. However, today, you see, some artists are still playing sold out.

With the advent of cybersensoriality, coupled with galloping individualism, we were told that we were all going to make love with ourselves, and never again in reality. This Valerian vision of the man and the woman seated in an armchair covered with electrodes never saw the light of day. When Google Earth arrived, the end of travel was predicted. It's true that now, you can walk in the Himalayas from your computer, but I don't have the impression that the airports are empty...

Yes, but right now it's barely fit for a child. It's dangerous, I think, for journalism, but for a novel, it's something else. ChatGPT does much better with two-page text than with 350-page text, with which there are other constraints, such as pacing.

Will ChatGPT be able to write an honorable novel in two years?

Probably. But no one can predict the relationship that the reader will have with such a text. We must not neglect the human relationship. ChatGPT will not be promoting his work, there will be no signing session! Remember when we started making very industrial food, remember the film L'Aile ou la Cuisse [directed by Claude Zidi in 1976, editor's note]... Well, today, everyone is looking for food organic !

Yes and no. I am convinced that Facebook, in particular, is one of the worst companies of the 21st century, and that the toxicity of social media is absolutely underrated. They have enabled a very small number of people to make phenomenal fortunes by impoverishing the population, and this is because the legislators are incapable of cultivating themselves.

They're not of that generation, they don't understand how it works, and instead of having the honesty to say "I don't know" and surrounding themselves with people who can provide non-partisan analysis, and to inform them properly, they made believe that they knew and their level is appalling today. Just look at the lamentable questions posed by the US Congressional Committee to the boss of TikTok, Shou Chew…

Yes, and so much so that I believe that, unlike the arrival of Facebook, today there is a collective awareness of the race against time and the danger of the consequences for society. We now know that we cannot count on those who manufacture the AIs. Those who develop don't think. Neither Google, nor Apple, nor Facebook are thinking. Otherwise Snapchat wouldn't have added an AI, which to me is criminal.

It's like putting scooters on the market without knowing how many have brakes. For them, the urgency is to put the scooter on the market, we will see later if it has brakes. The societal danger of artificial intelligence will very much depend on the courage of legislators to hold those who manufacture accountable.

In the agri-food world, an industrialist who poisons his consumers goes to prison. The manufacturers of the digital revolution must be subject to the same rules as those in the food industry. And on the other side, consumers should understand that what they put in their head is just as important as what they put on their plate!