[tnecio]

So true. And another benefit would be that domain experts giving a course could focus on teaching and sharing their knowledge instead of being forced to deal with all the organisational fluff around final grading and "catching cheaters" that is a giant waste of their time. (I see only usefulness in grading as a feedback mechanism for students – but not as "certification" of student's knowledge for the outside world. I also believe it would be more healthy for both students and teachers if you the grades were just a guidance tool, not something that will affect your future prospects at life).

At the end of the day the final grades from school / college grades depend on so many factors that this signal is close to noise anyway, but in college it often feels somehow more important than the actual learning and so much time and stress is spent on them.

In a better world I imagine it would be the organisations that need specific knowledge co-sponsoring "exam centers", separate from colleges, where you could go and get a certificate saying how well you know a given subject. Private companies that want to hire the best people actually have a good incentive to make these exams as fair and useful as possible.

To make an analogy with GAN networks in deep learning: the college would act as a generative part and "exam center" would be the discriminative part. It seems to work pretty well in ML, maybe it would work in education too?:D