Quantifying Underfitting and Overfitting: How much is "much lower" or "much higher"?

I understand the under/overfitting conceptually, how they are diagnosed and treated, but I don’t quite have a clear view of when they are really happening and when they are not.

In the lesson 3 notes we have this:

“… when our training error is much lower than our validation error [we have underfitting] … when your training set accuracy is much higher than your validation [we have overfitting]”

My question is: What is “much lower” and “much higher”?