Assistant Professor Zina Ward Collaborates on AI study

Why do large language models like ChatGPT overuse words like “delve” and “intricate"? An FSU philosopher and linguist have teamed up to try to determine the sources of these models’ lexical preferences. Using corpus analysis, a model comparison, and an online study, they explored the hypothesis that the culprit is actually the human feedback on which the models are trained.

Click here to read the article featured in Florida State University News.