Google's Quick Draw became five things.
- A pilot application
- A market test
- An experiment in massively data intensive reinforcement learning 1
- A data segue into handwriting recognition 1
- A data collection device for other branches of research (not necessarily an intentional purpose during conception) 2
Google's team on the project wished to delve into concept abstraction by tapping the minds of the authors of one billion drawings, attracting the public to the point of data acquisition using a contest between a massively scaled and parallelized game engine and the hundreds of millions of drawing authors.
The concept model (distinct from the algorithm and architecture) is a time series of motor actions. Due to the event model in computer peripheral devices and the coupling with the event model of the Netscape DOM, AJAX RESTful transactions can be stored into a huge data set as pen-up, pen-down, and pen-move operations. This is very much like some of the first photo-plotter commands pioneered by Gerber Scientific Instruments in the 1980s.
Learning Convergence Approach
The concept model is embodied within and trained using a generative recurrent network called sketch-rnn. 3
 Introducing the Kaggle “Quick, Draw!” Doodle Recognition Challenge,
 The Dead-Serious Strategy Behind Google’s Silly AI Experiments, by Katharine Schwab, 12/1/2017, FastCompany.com
 A Neural Representation of Sketch Drawings, David Ha, Douglas Eck, 2018