Caleb Schultz Kisby

Welcome! I'm a PhD Candidate in Computer Science at Indiana University, co-supervised by Larry Moss and Saúl Blanco. I'm also a member of the IU logic group and plwonks, and have participated in the IU Cognitive Lunch seminar.

Updates:

Contact:

Research

I study the foundations of machine learning and cognition using tools from logic and theoretical computer science. My current research focuses on questions that underlie neural networks and learning algorithms, especially:

• How should we best integrate symbolic and neural (sub-symbolic) systems?

• How can we extract, interpret, and verify the internal beliefs of neural networks?

• How powerful and reliable are different learning algorithms, when compared to one another?

• Is provably correct AI alignment possible?

I'm a mathematician by training, so I try to answer these questions by formalizing the underlying ideas and proving things about them. My particular approach to formalizing neural networks and learning algorithms involves tools and techniques at the intersection of dynamic epistemic logic, finite model theory, learning theory, and complexity theory. But I also try to embrace the inderdisciplinary nature of these questions, and apply ideas from philosophy, psychology, linguistics, and neuroscience.

News and Activity

This website was made using TeXmacs, based on Massimiliano Gubinelli's personal website.