To get a PhD, you typically have to enroll in a graduate program. Then you complete a few relatively easy courses.
You might need to pass a comprehensive examination that checks whether you have a basic understanding of the field. A few students fail at this point, but not many.Then you write a thesis and defend it.
In theory, you could write a strong thesis and still fail the oral defense. That is uncommon. The defense is typically a public event, and there may be guests. Failing a student at that stage would be a public humiliation. I have seen students who could not answer basic questions still receive their PhDs because the thesis itself was good enough.
Thus, to a very good approximation, getting a PhD amounted to writing a thesis.But we have a problem. AI can write something that looks quite a bit like a thesis. With a bit of prompting, you can produce something that looks like a PhD thesis.I have been telling everyone I can that we have a big problem.Apparently other people have realized this as well. People at Harvard are now saying that a thesis is insufficient for a PhD.
« The dissertation has in the past frequently been used as a proxy for the kind of mathematical development of a student that we expect: acquiring mathematical knowledge and demonstrating independent achievement. However, with modern AI systems, dissertations are not (…) a reliable tool for evaluating students, and PhDs should not be awarded primarily on the basis of the text of the dissertation. We thus recommend regular, multi-faceted evaluations of students in person, by multiple faculty, as the central component of assessment. These evaluations must be rigorous, not pro-forma, and should produce reports that might become part of the student’s dossier for future employment. »
Published by
