Montpellier launches AICET, the first AI assessment test to deliver the first AFNOR certifications

At the start of 2024, the University of Montpellier, the Metropole of Montpellier and the companies Numalis and BionomeeX are launching an AFNOR Spec to offer the first standardized artificial intelligence skills test. Named AICET for "Artificial Intelligence Competence Evaluation Test", it meets the growing needs of companies faced with the rise of generative AI and the arrival of European AI regulations (AI Act). This work is made possible thanks to the cross-fertilization of scientific and technological expertise, vision of uses and standardization know-how.

The AICET is a French and European response to one of the major challenges of the current AI revolution: to acculturate and train as many people as possible, in all sectors and at all levels, to understand the major challenges of AI. These challenges are manifold: regulatory, theoretical, technological, ethical, environmental... They require different levels of mastery to be taken into account.

 This approach is unique in France and Europe. Designed to be used both within companies and for training students and teachers, the AICET is intended to be a widely deployed certification after an implementation phase in Montpellier, and to be used, for example, for recruitment in all sectors. Designed to be a living tool, AICET will be added to over time, in line with the rapid technological evolution of AI.

Artificial intelligence skills: very strong needs and the need to be able to measure levels in different areas

The test is divided into 3 levels of expertise (acculturation, advanced user, expert) and 5 skill categories.

  1. Theoretical focus: fundamental knowledge of AI
  2. Applications: knowledge of AI application fields
  3. Operational focus: AI tools and methods
  4. Legal and ethical issues: laws, standards and regulations governing AI
  5. General knowledge: history and personalities of AI

"All research laboratories and companies are increasingly in need of AI skills. By providing a means of motivating learning, assessing skills, monitoring progress, and identifying experts and potential trainers, this test is of great importance. It will be a crucial tool for supporting the increase in the number of training courses offered to our students, staff and lecturers, as well as for offering training and certification to other audiences", emphasizes Philippe Augé, President of the University of Montpellier.

 Already tested both on students at the University of Montpellier, at the Polytech Montpellier engineering school, and on people undergoing retraining, this test is one of the pillars of Montpellier's Artificial Intelligence projects. Gabriel Krouk, Director of Research at CNRS, inventor of the AICET concept and scientific leader of the vIA-UM project in response to the "IA Cluster" call for expressions of interest, confirms: "The idea for this test came very naturally while working on the ambitious Montpellier project. It's essential to have this type of tool.

A standardized response based on Montpellier's expertise: AICET aims to become the "toeic of AI".

An AFNOR Spec-type reference standard will define a standardized test structure for assessing a person's competence in artificial intelligence, as well as the way in which this test can be completed and developed over time. The AFNOR Spec standard will be national in scope, and will then be proposed at European level to meet one of the requirements of the recently adopted AI Act, and then at international level. In this respect, the standard will draw on the expertise of Afnor and Numalis, who have already been involved in this type of project at ISO/IEC level.

A call for expressions of interest is being organized throughout January 2024, with a launch meeting scheduled for February 5, 2024. The standard will be developed on the basis of successive contributions from participants.

Its objectives are to :

  • Define the structure of the test and the main skill areas to be covered
  • Define the nature and distribution of question types and content-related constraints: proportion of MCQ-type questions, for example.
  • Define modalities and constraints for validating test questions and answers
  • Define the procedures and constraints for individuals and legal entities authorized to administer the test
  • Define the modalities and constraints of the individuals and legal entities authorized to mark the test and attest to an assessment level
  • Define the procedures and constraints for natural and legal persons authorized to issue certification based on a test.

More information on the AFNOR website.