Respond faster and more efficiently to users of public services thanks to artificial intelligence. That is the goal of the government, which will launch a six-month experiment at the start of the school year in September, he told AFP on May 11, confirming information from BFM Business.
Hence this experiment, which will train three generative AI models (ChatGPT, LLaMA and the French Bloom) to answer user questions, before a hypothetical generalization to the entire administration in 2024.
200 volunteer agents
“The content generation models will inevitably be wrong, we will have to train them little by little, to perfect them, in order to have the most accurate response possible,” explains the ministry. Each response will be subject to the control of an administration agent before being communicated to the user.
To carry out this experiment, the government hopes to find 200 volunteer agents. The AI is “a help” for officials, “not a substitute” that would allow the workforce to be cut, insists the Government.
“The idea is not to shake morale”, he still seeks to reassure the Ministry of Public Transformation, highlighting that AI is already used in Pôle emploi or in the tax administration, for other applications than the relationship with users.
An experiment carried out in collaboration with the CNIL
Given the confidentiality of certain data that citizens exchange with the administration, the experiment should finally make it possible to “eliminate” the problems posed by AI “in terms of regulation and protection of personal data.”
Therefore, it will be carried out in collaboration with the Cnil, the French data protection police and the national AI coordinator Guillaume Avrin.
Of great technical complexity, artificial intelligence systems, capable of writing theses, poems or translations in a few seconds, fascinate as much as they disturb. Scientific personalities have thus called for a moratorium on the development of the most powerful systems, until they are better regulated by law.
Source: BFM TV
