The honeymoon is already over. After arousing tremendous enthusiasm, the capabilities of generative AIs are beginning to cause concern. Italy and Canada have temporarily banned ChatGPT, suspecting that the chatbot violates regulations protecting personal data. And a thousand people, including Elon Musk, have called for research on generative AI to be put on hold for six months, the time to better understand the impact it could have on society and on employment. Interview with Corine de Bilbao, president of Microsoft France and Frédéric Bardeau, president and co-founder of the Simplon school, which created a training program dedicated to artificial intelligence in partnership with the American tech giant.
L’Express: The possibilities offered by generative AIs such as ChatGPT amaze as much as they frighten. In the translation, in the illustration, some have the impression that the AIs are going to steal their profession. Are they right to worry?
Corine de Bilbao and Frederic Bardeau: In many cases, trades will be transformed rather than fully automated. There are many things that AIs are not able to do as well as us. Often what is simple for humans is complicated for AI and vice versa. This is called Moravec’s paradox. Moreover, generative AIs do not understand what they are doing, the meaning of the sentences they write. Rather, they should be seen as assistants. The better they are briefed, oriented by the demands of a human, the more they will give good results. Who better, for example, than an experienced translator to bring ChatGPT to produce the finest possible translations? By delegating part of their missions to AI, individuals will be able to focus on other social and creative tasks, with higher added value. For example, lawyers will spend less time sifting through legal texts to find specific details and more time discussing with their clients and preparing their arguments. Generative AIs are finally removing the barriers that prevent us from going into certain areas. Those that generate code, such as the Copilot GitHub, for example allow Internet users to develop certain tools without necessarily mastering a programming language.
Teachers worry about students using these tools to cheating on homework. This has already happened. How to adapt the educational world to this new deal?
In the same way that young people should not be banned from using Wikipedia, ChatGPT should not be banned, but they should be taught how to use it. The arrival of calculators, the Internet and GPS apps have made certain skills superfluous, but new ones immediately take their place. For teachers, generative AIs can be interesting tools that will reduce the time spent on certain tasks (preparing information summaries, etc.) in favor of other more strategic ones, in particular supporting students or debating with the class.
Digital has automated many tasks. Will the switch operated by the AIs be different from what we have already experienced?
There have, in fact, already been similar concerns in the past, especially since vague but alarmist studies are regularly appearing. The difference of generative AIs is that they will not only automate low-skilled jobs. What a recent study by OpenAI shows is that they will also transform many skilled trades, especially at bac + 3 level.
How to train for this new emerging economy?
All professionals are going to have to learn how to use AI, to ask it the right questions to get the right results. Creatives should learn to use Midjourney, and ChatGPT secretaries. It must also be shown that AI professions are not reserved for bac +5 and bac + 8. The sector has a huge need for intermediate level AI developers. People who will not create an AI from scratch but will be able, for example, to develop apps that connect to existing AIs and who know how to manipulate the data on which these AIs are made to work. Many people censor themselves and do not dare to go into the field of artificial intelligence, because they think you have to be a math genius to work there. The training courses of the Microsoft by Simplon IA schools, which have just celebrated their 5th anniversary, of course require effort. But we train very varied profiles: people who don’t have a baccalaureate or who have master’s degrees in fields far from tech, such as law or psychology, and are in retraining. The only prerequisites are to have the mathematical level of a final year student and notions of programming because the duration of our training does not allow you to start completely from scratch. Our insertion rate is 98% and that’s not surprising. The sector has a great need for these profiles. In the field of data and AI, there are around twenty thousand vacancies in France.