Comment by jkrshnmenon

I'm also trying to understand the implications of this work.

Does it imply that some for some functions F(x) = y, you can compute x given the value of y without computing the inverse of F ?

If so, what constraints does F need to meet for this ?