La Poste GPT, the first generative AI for postmen, and the new ethical AI charter – ZDNET


The big French technology fair VivaTech opens its doors in Paris today. For the La Poste group, this meeting is an opportunity for two announcements in the field of artificial intelligence, where it has a total of almost 100 projects.

Already equipped with a data charter since 2016, updated in 2021 to anticipate the entry into force of the AI ​​Act, the company now has a new data and AI charter. Why this update? In order to strengthen the ethical design of your AI projects.

Five of the external members of the Trusted AI Committee

The document is not intended to be a simple declaration of general principles. La Poste thus emphasizes that its charter defines the “operational framework of the activity”. The aim is to guarantee “reasonable use of data and ethical design of algorithms”.

To ensure compliance with the established ethical principles, La Poste and its AI teams will rely on several control mechanisms. Among them, the Committee for Trusted Artificial Intelligence composed of 23 people, including 5 external ones.

Françoise Soulié-Fogelman (scientific advisor of the French IA Hub), Vincent Roca (researcher, head of the PRIVACTIS project at INRIA), Éric Dadian (president of the AFRC), Célia Zolynski (law professor and coordinator of the Observatory ‘ The IA sits on this committee) and Eric Salobir (President of HTF).

Trusted AI officers deployed

The main mission of its members is “in particular (…) to support project teams in determining the purposes of data processing and to guarantee an ethical process in the development of projects.”

However, such a device is not enough. So La Poste specifies that it will provide its teams with pre-filtering and self-assessment grids for AI projects. From the beginning, the initiatives are analyzed according to four criteria (possible purposes of the algorithm, legal regimes applicable to the data, technical risk factors and finally users and potential beneficiaries).

The result of this investigation will then be analyzed by trusted AI officers combining business and GDPR officers. La Poste is gradually deploying trusted AI officers internally to support its AI designers.

200 testers for La Poste GPT

In addition, the company provides awareness training to its 230,000 employees. Participants in its data career training courses have access to “advanced training” on ethical issues.

La Poste also states that these principles apply to all artificial intelligence, including generative. In this sector, the group is currently working on the development of its first IAGen tool called La Poste GPT.

The solution is presented at VivaTech. La Poste GPT is still being tested. Among employees, more than 200 across sales and customer relations teams now have access to it. The company says it’s considering a phased rollout by the end of 2024.



Source link

Leave a Comment