Laboratory on Autonomous and Empathetic Multimodal Agents – LAMAE

For several years, the Carnot Interdisciplinary Laboratory of Burgundy, UMR CNRS 6303, has been developing research on generative AI and the use of this AI in business processes. For example, printing 4D objects made of smart materials is a real challenge in terms of designing and understanding the behavior of smart materials over time. Added to this is the effervescence of the creation of new materials, complicating the design process. In addition, recent scientific and technological breakthroughs in the field of generative AI are opening up new perspectives for the creation of tools to support a specialist in his or her business process.
Thus, the use of large language models (LLMs) as a central pivot in the design of autonomous agents represents a promising prospect in the field of artificial intelligence. Proofs of concept such as AutoGPT, GPT-Engineer and BabyAGI provide inspiring illustrations of this approach. The potential of LLMs goes beyond simply generating sophisticated linguistic content to encompass an ability to act as a generalized problem solver. The agent can break down complex tasks into sub-objectives, reflect on previous actions to improve, and master the use of external programming interfaces to access additional data. Within an autonomous agent system guided by an LLM, the LLM assumes the central function of agent intelligence, in synergy with various critical components, including scheduling, short- and long-term memory, as well as the use of external tools. This revolutionary approach opens up new perspectives in the development of autonomous agents with effective adaptation and execution skills in the face of complex cognitive and operational challenges.
An "autonomous digital human" refers to a virtual representation of an individual, used in various digital contexts such as virtual reality, video games, communication applications, etc. This representation can take the form of an avatar, a virtual assistant, or an interactive character. DAVI, a long-standing partner of the project leader and committed to this joint laboratory proposal, is a major player in the field of emotional conversational agents. Their agents have the ability to interact with users in an empathetic manner, making interactions feel like human interactions. The objective of the LAMAE Labcom is to develop knowledge in the field of Autonomous and Empathetic Multimodal Agents while implementing innovative and valuable technologies for the creation of digital humans in promising markets. We aim to create digital humans capable of interacting in a natural way with users to improve the interaction experience over the long term by integrating the relational dimension, unlike chatbots developed only to answer questions. We aim to develop agents with the ability to perform complex task sequences in a variety of areas, to answer business questions, to train, but also to keep their business knowledge up to date, while adopting proactive behavior. The focus would be on the design of advanced natural language processing, computer vision algorithms to create personalized and immersive interactions.