There's an episode of Star Trek TNG where the crew members were being abducted from their sleep by aliens who performed medical experiments on them but attempted to wipe their memories so they'd never know. However, enough memories lingered disguised as dreams that when the crew started talking amongst themselves about why they all felt exhausted instead of well-rested, they realized they were all familiar with the bits and pieces of each other's dreams. Seeing now that it must be real, they put their combined memories together to form a picture... Using the holodeck of course. They all stood in the empty grid of the holodeck and gave the computer verbal commands to create objects and sounds as they vaguely remembered them to recreate the room into which they had all been abducted.
In my dream, Google had launched a new service. It featured a line-drawing sort of output window that would draw a vector-based scene based on a description that you type into an adjacent text box. I specifically remember seeing a picture of a red racecar being unloaded from its transport truck with two of the pit crew members in the scene. I'm not sure if that was a scene rendered by the service or if it was a normal photo, but somehow it became my task to recreate that scene using this Google service. So I began forming one long, descriptive, complete sentence to describe the scene, and with every word I typed the output window would draw something new or modify something it had already drawn, and I was amazed at how well it resembled the original scene that I was describing.
Tell me that's not something Google will develop in the next 5 years. :)