@Kitonum 

Kitonum's answer works best for me.  I want to use 1.5^x not (3/2)^x.  Thanks Kitonum.

Regaridng the comment "If you want to do symbolic computation, don't use floats."  Rather it seems to me this comment means "It's hard to force Maple to do symbolic computation with floats, and so it's better not to try."  The first half of this statement is true, but I want to try.

Why do I want to try? I'm using Maple to generate 20 random derivative problems for my students.  I want the Maple generated problems to look the same way as the problems in class and in the book: some of them have things like 1.5^x.  In any case, why shouldn't I be able to write things like ln(1.5) without having Maple evaluate it?  And isn't this sort of similar to how people often need inert versions of procedures like Sum and Int, etc.?  Doesn't that show that a reasonable use case exists?