Drawing on a touchscreen or trackpad can be a huge pain – but when you’re on the go, sometimes that may be your only option to quickly convey an image. To the rescue, Google has unveiled its latest AI experiment, AutoDraw, which uses machine learning to pair your wobbly doodle with a corresponding artist-drawn image – like autocorrect for sketching.
The interface appears as a basic drawing program, with shape, fill, text and free-draw options to select within a menu on the left of the screen. At the top, the “AutoDraw” button is the true novelty. Start drawing, and the program will begin making guesses as to what you’re attempting to portray. Tapping on one of these options will replace your sketch, and then give you the ability to resize or move it around the page. For us architecture enthusiasts, AutoDraw currently includes representations of some famous world landmarks: the Eiffel Tower, the Great Wall of China, One World Trade Center, to name a few.
AutoDraw will also help correct your simply geometries, resulting in quick ways to draw perfect cirlces, triangles or squares. As much as architects appreciate the natural stroke of a pen, we also appreciate precision. Using AutoDraw to sketch out some geometric concepts could help make the diagram as clear as possible.
What other implications could this have for architecture? While the technology is currently limited to the selection of preloaded images, it’s easy to imagine that in the future, users could upload their own sets of buildings or entourage, allowing people to create quick drawings that still express the same language as their own personal drawing style. Further, a crowdsourced database of these buildings and figures could be a possibility, letting users compare their own drawings to those of people around the world.
For now though, the program is still a whole lot of fun to play with. Give it a try for yourself, here.