Neuroscientists Debate Future of AI Models Predicting Human and Mouse Brain Activity

Researchers are currently divided over whether foundation models designed to predict brain responses can truly explain how the human brain functions. Several well-funded teams, including Meta, Stanford University, and Neuralink, are investing millions to train artificial neural networks on vast amounts of brain data. This approach mirrors the development of large language models, aiming to create predictive systems for both human and mouse brains. Experts Bing Wen Brunton and Patrick Mineault are leading a debate on whether these <b>AI models</b> can effectively decode biological processes. They question if accurate prediction of brain activity is sufficient to understand the underlying mechanisms of the brain or if modeling the body is also required. These findings will be discussed during a live session at the Society for Neuroscience annual meeting in Washington, D.C., this Sunday.