The USC Center on AI Foundations for the Sciences (AIF4S) is seeking Summer/Fall graduate student interns focusing on three areas (see below). This is open to PhD students at US universities. Interns will have the opportunity to collaborate with various domain experts and will have access to unique data sets and computational resources. For full consideration please apply by March 12, 2024 using this form.
Mechanistic understanding & enhancing reasoning capabilities: This area focuses on understanding the inner working of foundation models (including LLMs, VLMs, and other multi-modal models), understanding their failure modes and how to fix them, data filtration/curation strategies, and strategies for enhancing their reasoning capabilities. In particular, our focus is on gaining a better understanding of the current limitations of pretrained visual representations in foundation models.
VLMs in Healthcare: This area focuses on reliable and trustworthy deployment of multimodal models including vision language models in various healthcare settings. We are interested in augmenting the workflow of medical professionals with advanced foundation models to bring down healthcare costs and improve efficiency while maintaining alignment between doctors and highly capable AI systems.
AI Foundations: This area focuses on developing theoretical foundations for AI in areas such as representation learning, feature learning, self-supervised learning, LLMs/VLMs, in-context learning, data filtration, and scaling laws and other emerging learning paradigms.