Deep language learning: from fiwGAN to LLMs

Abstract: 

Amount Awarded: $25,000

Artificial Intelligence is driving a paradigm shift across scientific fields, including linguistics and cognitive science. This project pioneers the application of state-of-the-art deep learning models to study how humans learn and use language, addressing pivotal questions in cognitive science, neuroscience, and language evolution—namely, which properties of language are uniquely human, and how they might emerge in artificial systems. Although these advanced modeling approaches have transformative implications for our understanding of the human mind, they are typically absent from standard curricula.

To fill this gap, our team will conduct a series of educational workshops at the University of Oslo and UC Berkeley, introducing students at all levels to cutting-edge deep learning frameworks for linguistic and cognitive modeling. By jointly training students to apply models to real-world corpora—such as a Norwegian language dataset—we will seed collaborative research that explores language structure, evolution, and neural correlates. Through these workshops, we will establish a cross-institutional research group, produce open-source training materials, and foster an interdisciplinary community to drive innovation at the intersection of AI and language science. In doing so, we will not only equip the next generation of researchers with modern computational tools but also advance our theoretical understanding of human language.

Author: 
UiO PI: Dag Haug
Berkeley PI: Gasper Begus
Publication date: 
July 1, 2025
Publication type: 
Grant (UiO)