General presentation
Artificial intelligence and data science have been at the heart of the LIRMM’s activities for several decades. The scientific and societal importance of these topics is undeniable today.
Our research teams contribute to the methodological aspects of symbolic and sub-symbolic AI, with a recent development of work in the field of statistical machine learning. They aim to bridge the gap between our well-established research activities in symbolic AI and the more recent work in numerical AI that is increasingly being adopted in our work. They also target the efficient and robust implementation of embedded AI thanks to emerging computing paradigms, architectures and technologies (neuromorphic computing, resistive memories, …). Finally, they study numerous applications of AI in many fields, such as robotics and microelectronics. Our teams also work on modelling and data processing, as well as knowledge engineering in relation to numerous application fields, for example the environment, agronomy and health. Current applications concern a wide range of fields, such as the environment, agronomy and biology (with CIRAD, INRAE and IRD), health (CHU, IDESP), remote sensing (Espace Dev, TETIS), human and social sciences, microelectronics and robotics.
Our specificities
– Trusted AI: explainability of black-box AI systems in general, argumentation and human/machine dialogue, robustness of hardware architectures dedicated to AI;
– Mathematical foundations of AI: logical approaches and algorithmic and complexity studies, dynamic neural networks taking into account the suppression of inductive biases via so-called “Transformers” models.
– Embedded AI and frugal AI: energy efficiency through the design and programming of low-power hardware accelerator architectures and neuro-inspired architectures. The architectures studied in this way target embedded systems, for example in a distributed context such as edge computing. On the other hand, we are interested in AI methods applicable to the parsimonious management of frugal data lakes storing masses of data while waiting for their exploitation.
List of teams involved in the theme
ADVANSE, COCONUT, FADO, GraphiK, ICAR, MAB, MAREL, SMILE, TEXTE, ZENITH, ADAC, SmartIES, TEST, DEXTER, IDH
Specific transverse action
– HAIR (Human, AI & Robotics). This is a transverse action, conducted jointly by teams from the Computer Science and Robotics departments. The objective of this action is to study the multidisciplinary aspects of human-robot interaction. Led by a computer scientist (M. Croitoru) and a roboticist with a background in neuroscience (G. Gowrishankar), the action regularly organises meetings between LIRMM members, has obtained INS2I funding to study artificial mediation by robots (Master’s thesis of S. Victor) and organises an international workshop each year (HRI@Montpellier).
Collaborations and visibility
– Local partnerships: IMAG (specialist in digital AI techniques), ESPACE DEV (working on ontologies and constraints), TETIS (specialist in geo-spatial data)
– Projects: ANR (https://www.lirmm.fr/projets-anr/) and EU (https://www.lirmm.fr/projet-europeen/)
M2 internships across teams and departments focused on AI and Data Science, funded or supported by the AI-DS cross-cutting axis.
— AI assistance for controlling a discrete snake-like robot – Benjamin Calmé and Yassine Haddab (DEXTER, MIC), Jérôme Azé (ADVANSE, INFO)
— Generation of program invariants using a large language model – Simon Robillard (MAREL), Maximos Skandalis (TEXTE)
— Few-shot learning with Large Language Models for Natural Language Inference tasks in French – Maximos Skandalis (TEXTE, LIRMM), Richard Moot (TEXTE, LIRMM), Michael Sioutis (COCONUT, LIRMM)
— Hybrid methods for recognizing textual inferences involving discourse markers such as “almost” and “barely” – Pascal Poncelet (ADVANSE), Christian Retoré (TEXTE)
— “Only ChatGPT gets me”: emotion detection and ChatGPT – Madalina Croitoru (IDH) and Konstantin Todorov (ADVANSE)
— Multimodal analysis of medical records and images using AI for automatic analysis of chronic wound evolution – Marc Chaumont (ICAR), Gérard Subsol (ICAR), Mathieu Lafourcade (TEXTE)
— Multimodal analysis of medical records and images using AI for automatic analysis of chronic wound evolution – Marc Chaumont (ICAR), Gérard Subsol (ICAR), Mathieu Lafourcade (TEXTE)
— Impact of physical embodiment in artificial agents for human-robot cognitive interaction – Madalina Croitoru (BOREAL), Ganesh Gowrishankar (IDH)
— Anaphora resolution and proof theory – Richard Moot (TEXTE), Simon Robillard (MAREL)
— Mixed Reality applied to 3D temporal biological data – Emmanuel Faure (ICAR), Nancy Rodriguez (ADVANSE)
— AI assistance for controlling a discrete snake-like robot – Benjamin Calmé and Yassine Haddab (DEXTER), Jérôme Azé (ADVANSE)
— Pizza or sushi tonight? A computational model for human/robot negotiation – Madalina Croitoru (BOREAL), Ganesh Gowrishankar (IDH)
— Enhancement of a prototype for fish tracking using Deep Learning from stereo image sequences captured by an autonomous underwater robot – Cyril Barrelet, Gérard Subsol, Marc Chaumont (ICAR), Karen Godary-Dejean (EXPLORE)
Animations
- Seminar “Behind the Laughter: What It Really Takes to Understand Memes” – February 4, 2025
Guest: Filip Ilievski, Vrije University, Amsterdam
Topic: Meme detection and analysis through neuro-symbolic methods - Seminar “Factuality Challenges in the Era of Large Language Models: Can we Keep LLMs Safe and Factual?” – November 18, 2024
Guest: Preslav Nakov, MBZUAI (United Arab Emirates)
Topic: Fighting misinformation with contemporary AI and NLP, AI and fake news detection - Seminar “Creativity and AI: Master Class by Markus Eronen and Interdisciplinary Round Table” – September 27, 2024, from 2:00 PM to 5:00 PM
Auditorium, St. Charles 2, Epsylon Laboratory, 2033 Av. Bouisson Bertrand, 34090 Montpellier
Inter-laboratory event supported by Montpellier Metropole
Guest: Markus Eronen (Univ. Groningen, Philosophy of Science)
Format: Presentation, master class, round table (moderated by Sarah Labelle – LERASS)
Involved laboratories: LIRMM, LERASS, LHUMAN, Epsylon
Topic: Generative AI & the nature of creativity, Creative credits, Definitions and measurement of creativity in AI - Seminar by Agata Gurzawska on Ethics and AI – Thursday, November 30, 2023, from 10:00 AM to 12:30 PM at LIRMM. Guest: Agata Gurzawska (Trilateral Research Ireland)
Topic: Integrating ethics and human rights into AI; Systemic approaches for responsible AI - Seminar “Hybrid AI: Applications and Challenges” – Wednesday, September 13, 2023, from 9:30 AM to 2:00 PM at LIRMM.
Guests: Achim Rettinger (Univ. Trier, Germany), Hybrid AI & Knowledge Graphs (Rettinger)
José Manuel Goméz Pérez (Expert.AI, Spain), Deception Detection & AI (Pérez)
Round table moderated by: K. Todorov
Discussions: Toward a Franco-German Horizon / ANR PRCI project - International workshop within the transverse action HAIR: “AI for improving machine interactions with humans and the environment”, 27 – 28 October 2021, Montpellier https://sites.google.com/view/hrimontpellier/home