A lot of highly specialized human knowledge lives in PhD theses and dissertations that are never broadly digitized, never fully published, or are difficult for frontier models to access. I was watching a tiktok about a woman loving the concept of social media because it allowed her to share her thesis on children’s jewlery through the ages. I think there may be an opportunity for OpenAI to build an opt-in, rights-cleared acquisition pipeline for this kind of material.
The idea would be to invite PhD graduates, researchers, and possibly universities to voluntarily submit theses/dissertations and related supporting materials under explicit licensing terms, with compensation for access.
What seems especially valuable here is that this would not just add more text. It could add expert-dense knowledge in sparse domains where only a small number of people may have deep expertise, and where important synthesis, observations, or edge cases may never have made it into broadly available publications. I could list out a variety of usage cases but i’d rather leave it to the imagination.