Julian Michael

I am a postdoc at the NYU Center for Data Science working with Sam Bowman. I earned my PhD from the University of Washington where my advisor was Luke Zettlemoyer.

I work on a variety of things, mostly focused on AI alignment or formal semantics of natural language. In alignment, I focus on scalable oversight and agent alignment, from the lens of task formulation, data collection, and evaluation methodology. I’m especially interested in using debate as a training and evaluation paradigm; see our recent paper. In language, I work on ways to design, annotate, and model semantics in a scalable, data-driven way while taking advantage of our understanding of linguistic structure; see my thesis summary paper. I have worked on approaches to crowdsourcing annotation for syntactic parsing, semantic role labeling, and predicate-argument structure.


I have been involved in the development of QA-SRL and related formalisms, where I build systems for annotating and modeling linguistic structure at large scale. I have released two corpora:

For QA-SRL, we have an online explorer for the dataset. Code for manipulating, crowdsourcing, or browsing QA-SRL data is available here. Recently, I have shown that QA-SRL can be used to recover traditional discrete semantic roles.

I am also interested broadly in the Science of AI and NLP, using empirical methods to improve our understanding of intelligent behavior and language use. Along these lines, I have worked on broad-coverage and fine-grained evaluation of models, unsupervised discovery of linguistic structure, and explicitly incorporating ambiguity into task design. See my publications for a full list.

Random Thoughts