Listen to our podcast Cogitons Sciences: For an ethical and regulated integration of AI [L’IA s’invite dans nos métiers #3]

2024-04-02 06:00:23

Welcome to the 21st episode of Cogitons Sciences, the podcast that deciphers science issues! The third episode of this mini-series invites Jean-Gabriel Ganascia, professor of computer science at Sorbonne University and president of the CNRS Ethics Committee (COMETS) to speak on the ethical and regulatory aspects of the integration of AI In the enterprises.

To understand the ins and outs of the European regulation on artificial intelligence, we interviewed Jean-Gabriel Ganascia, computer scientist and philosopher specializing in artificial intelligence and ethics at Sorbonne University. He is also a member of the CNRS ethics committee. He also shares with us the right questions to ask before developing AI.

Listen to the episode here !

What is the AI ​​Act? [01:32 – 16:54]

The European IA Act regulation aims to “to protect ourselves once morest the dangers linked to artificial intelligence” explains Jean-Gabriel Ganascia. And this, by introducing the notion of risk which is “the possibility of danger” he adds. These risks are categorized: unacceptable (subliminal techniques, scoring, real-time biometrics), high, moderate and low (spam filters, video games).

For the moment there was little regulation specific to each country, and even less in France. Reason why this regulation has become important. In the United States and China, the framework is very different. In the United States, since July 2023, regulations have required companies to develop AI for the common good. In China, “the ethical principle on which the reflections are based is that we must first ensure security and cohesion” says the expert.

Worried businesses [17:04 – 21:14]

Many companies are concerned regarding the complexity of the regulations, with general and restrictive rules, or even exceptions to the rules. In other words, “we need competent legal services who have a perfect command of these regulations to guide […] and for small companies, it is extremely difficult” regrets Jean-Gabriel Ganascia. The regulation also included generative AI late, and urgently, particularly with the arrival of ChatGPT on the market.

Integrating ethics into development [21:18 – 31:46]

Should we ask questions regarding ethics before embarking on the development of AI? For our expert, questions must be asked following mastering AI techniques but before implementing projects that use it. “On each project, we must ask ourselves what are the consequences and deviations in the use of the system” specifies the president of COMETS, following a precise methodology in four points: assemble an ethics committee, ask ourselves what constraints we will impose on ourselves to avoid misuse, have it certified by an independent body and keep supervision. And if you don’t know how to put all this in place, there is training!

Click here to listen to the episode !

References cited:

Resources to go further:

Cogitons Sciences is a podcast produced by Techniques de l’Ingénieur. This episode was directed by Séverine Fontaine, in collaboration with Marie-Caroline Loriquet.

1712103045
#Listen #podcast #Cogitons #Sciences #ethical #regulated #integration #LIA #sinvite #dans #nos #métiers