The search engine giants of Google have been dabbling in the artificial intelligence technologies; recently they have been making huge steps in the progress of computer techniques. Software named Google Brain, creates techniques of a computer vision that will allow software companies to bring in content of hand drawn images allowing them to re-create those on the fly. This would allow a collaboration of drawings, doodling, or artwork to be connecting with a neural network of AI. That will then proceed to the best of its abilities using predictions and past experience to digest millions of user generated samples.
Remember when predictive text first came out? At first it was not perfect but then over time it merged into the idea of auto correct, based off words you normally use or misspell. Now it generally corrects your mistakes with ease and so smooth that at times you did not even notice the correction-taking place. Now, take that same idea and spin it into picture and media!
Sketch-RNN which was introduced in early April, suggests the neural net is gaining intelligence quickly based off user made doodles and drawings. Sketch-RNN, will attempt to guess the next part of your sketch. For example, if you are thinking about drawing a dog or a cow then using the neural net technology it would attempt to make the second part of the drawling. So you begin with the outline of the body of the dog, Sketch-RNN then will draw the eyes or nose, potentially the ears or the legs. Then you begin to draw the next and so on and so forth until the drawing is complete.
This gave the idea called Quick Draw, where human users are continually trained to teach the AI what it should be learning. Now the idea seams simple, but if you really dig into the idea you would realize that you could easily draw something in mere seconds rather than hours. Even better yet suggest images or gif’s for you media messages instantly instead of looking for them, the implementation of this technology is limitless!
Stop and think about anything from art work to the statue of liberty, Sketch-RNN at the moment can be pretty hit or miss but once this intelligence gets closer to perfection it will get smarter and smarter making our lives a bit more convenient. Multiple Predict is a program within the Sketch-RNN that will product several versions with different outcomes from the project drawn. This process helps them predict the images that could be completed and help it learn what the human brain is thinking when starting to draw or doodle! Making a crude drawling with the computer mouse just may become quicker and more efficient along with much nicer product outcome at the end!
The higher arching idea is for these programs to improve over time as more people use the program and keep feeding the AI learning portion so that it can gain the data it needs. Creating an end goal to teach Google’s software to visualize real life objects and then repopulate them digitally with a detailed understanding of how the human thought creates images. Making it quicker and more efficient to create data rich items for projects or media messages.