In Lesson 11 at 58:40 Jeremy argues that having more output channels than the number of dimensions is waste of computation. Is that so?

If I have a problem where I only have one number as input and for predicting the target it is useful to have such information as “is the input greater than or less than ‘a’ and greater than or less than ‘b’?” than I have a problem where the number of dimensions of the input is 1 and the number of output filters I have is 2. And so I have a problem where Jeremy’s argument doesn’t hold, therefore the argument is flawed. Am I missing something?

In Lesson 12 at 30:13 in XResnet, the number of input channels is 3 and the number of channels of the first 3x3 kernel is 32 and the number of dimension of it’s input is 3x3x3=27 < 32.