Aix-Marseille Université

AI-based models are ubiquitous in many scientific domains, with fast-evolving technologies and actors, and M/EEG research is no exception. On the one hand, transformers offer powerful models for brain time series encoding/decoding, even though foundation models have yet to be largely adopted by the community. On the other hand, AI tools are currently revolutionizing experimental practice, offering an impressive range of empowering agents for M/EEG practitioners. But a good MEEG practitioner should know the tricks of the trade for all the tasks required to complete a study: experimental design, bibliographical search, code writing, results interpretation, paper writing… and large language models (LLMs) — such as ChatGPT and other conversational agents — are available to assist you in all those tasks. Should you use them? Could you have confidence in the bibliographical references of LLMs? Is it possible to do vibe coding for your preferred MEEG toolbox? And what is vibe coding and could you do vibe research? During this lecture, we will investigate all these topics and more, like the environmental cost of using LLM, their legal implications and their ethical limits.