Have you ever gotten an A on a test, and then realized later that you forgot everything you memorized? Why is that?
It's because our tests are hackable. In the essay "The Lesson to Unlearn", Paul Graham says that the most damaging thing students learn in school isn't any particular concept, but rather it's learning to get good grades.
What can we do about this problem? The solution is to make better tests, that actually look at if you deeply understand a concept. Is there already a kind of test that does this effectively?
One answer is to use oral exams, because they allow the examiner to ask why, and really see if the student understands things from first-principles.
Oral exams are expensive and time consuming to administer — what if we could automate it using large language models? This is what I'm trying to create.
Request to professors: Send me your lecture notes and I will train a bot to have a Socratic dialogue on specific units of the course.
Fill out the brief interest form below. I'll follow up and send a link for you to play around with it as soon as possible.