by Laurie Sullivan @lauriesullivan, April 12, 2017

Google has created several experiments using artificial intelligence (AI). The latest: a Web-based drawing tool it calls AutoDraw, which turns simplistic doodles into comprehensive pictures. Think of the tool as a combination of the Google autosuggest feature in search and Microsoft Paint.

Doodle on the desktop or mobile site and the suggestion tool will start guessing as to the type of image the artist wants to draw before providing similar options to complete the picture.

The AI software offers a pallet of tools, such as Select, Draw, Type, Fill for colors, Shape, and a pencil for AutoDraw.

Google wants to make it easy for anyone to understand and start exploring AI. It connects AutoDraw to a site called AI Experiments where it showcases several experiments and asks developers to create and submit their own.

Several have been submitted, such as Transit Flow by Ray Luong, a product designer at Inkling, though he says the project didn’t actually use AI. Transit Flow is an experiment using technology and data from the Northern California massive transit system BART to explore how the flow of rider traffic across the Bay Area transit system changes throughout the day.

Luong has lived in the Bay Area for more than five years and has always been interested in the people and stories that make up the urban landscape. “After seeing the climb in BART ridership over the past few years, I thought it would be an interesting narrative to visualize the pulse of people coming in and out of the city,” he told Search Marketing Daily. “For me, it’s fun watching the city wake up around 6 a.m., then imagining thousands of people going home after work.”

Google’s site includes open-source code and resources to help developers get started. The code includes Google’s APIs for computer vision and speech, along with TensorFlow, Google’s open-source library for machine learning that lets developers build deep neural networks that make AI work, and much more.

In the early days, everything was built on logic, doing mathematical integration problems, such as playing chess. Now, developers realized hard, logical rules are not the way to solve everyday problems, like identifying similar pictures or completing a drawing.

Unlike the early computers and software, these types of problems are not black and white, but nuanced. Google has set out to test several options, such as pattern recognition, reinforcement learning and probabilistic machine learning, to address these issues.

In addition to AutoDraw, Google advances in AI have surfaced in experiments such as machines against human players at the board game Go.

For AutoDraw, Google used the technology behind Quick Draw!, an AI experiment that tells the drawer what to draw and tries to guess the doodle created.

MediaPost.com: Search Marketing Daily

(36)