Multidisciplinary Institute in Artificial intelligence
For this webinar, we will present 2 use cases involving generative AI:
1/ 3D Creation and Generative AI: Challenges of a New Dimension
The goal of this presentation is to introduce a novice audience to the potential of generative artificial intelligence for 3D. We will start with an overview of existing generative AI technologies and applications, before detailing the various 3D generation techniques available today.
This presentation aims to inform attendees about the state of the art in this technology and its future developments. We will demonstrate how generative AI for 3D is already operational and holds significant disruptive potential for numerous industries. Finally, we will conclude with a discussion of the current limitations of this technology and future research perspectives.
Speaker
Glenn Avezoux, Co-founder and CTO, Chat3D
This talk will propose an overview of the current research and opportunities of Code Language Language Models (CodeLLMs), an emerging area of research and development at the intersection of natural language processing and programming languages.
We will present the opportunities that CodeLLMs propose, such as their potential to advance automatic code comprehension, generation, and refactoring, as well as the challenges that remain to be addressed, including robustness, generalizability, alignment and evaluation metrics.
We will also highlight recent advances in this field, handle in my lab, with a particular focus on three areas of research.
First, we will discuss the intersection of formal verification and CodeLLMs, exploring how these models can be leveraged to benefit from formal specifications and allow for the development of verified software.
Second, we will examine the use of Retrieval-Augmented Generation (RAG) techniques with CodeLLMs, which can significantly enhance the quality and relevance of generated code by integrating retrieval mechanisms from a large code corpus.
Lastly, we will touch upon the application of CodeLLMs in vulnerability detection. By understanding and leveraging the capabilities of CodeLLMs, we can potentially automate the identification and mitigation of security vulnerabilities in software, thereby enhancing the overall resilience of our digital systems.
Speaker
Julien Perez, Associate Professor
Laboratoire de Recherche of l'EPITA - AI Team
