What were the problems with GAN's , How did they solved?

As this is deep learning forum. I thought to ask this question anyway.
As Ian talked about GAN’s problems. About that weird structure GAN were making given in the Nips lecture.
As GAN today progress much much better.

Problems he mentioned was

  • Problems with counting
  • cherry picked
  • global_structure
  • perspective

as these problems gave some images like
counting global structure Perspective cherry_picked

What conditions because of that they got this problem, how was it solved ?
If you know , please tell each solution of problem for mentioned problem.
Thank you for help.