Yejin Choi

Yejin Choi

A researcher in natural language processing, formerly at the University of Washington and the Allen Institute for AI and now at Stanford, whose work targets the commonsense knowledge that language models absorb unevenly: the everyday physical and social expectations people never write down. Her projects include the ATOMIC commonsense knowledge graph, the COMET models that generate such inferences, and Delphi, an experiment in machine moral judgement that she used to expose the limits of the approach as much as its promise. She received a MacArthur Fellowship in 2022. (See also: Large language model, Hallucination, AI alignment)