A criterion for automation versus augmentation in AI

Ask every person about their dreams. List all human activities needed to make those dreams happen, which will likely be only a subset of all current human activities. Augment the dreams. Automate all the direct and indirect dependencies that are not part of anyone's dreams. Ignore the rest.

…

— We've found the cancer cure.
— Great. Time to design reasonable steps to subtly guide the human group closest to the solution to discover what we have.
— Is it really necessary? Can't we publish it?
— Remember our agreement: we discover it, but we must make them think they did it. They can't live without it; they call it a meaningful life.
— Let me see… well, there is this retired woman doing experiments in her garage, because she wants to save her sick child.
— That's perfect, as we must also optimize for future movie plots about their discoveries.