Featured Image for This Google AI can guess what you’re drawing, and it’s getting smarter
Apps

This Google AI can guess what you’re drawing, and it’s getting smarter

Can a neural network learn to recognise doodling? That’s the question posed by Quick, Draw!, an online game from Google built with machine learning.

It’s simple, really: you draw an object, and a neural network tries to guess what it is.

For those unfamiliar with machine learnings, it’s basically an application of artificial intelligence that allows it to learn from experience without explicit instruction or human input (via programming).

Quick, Draw! has been around since last November, but Google has just released a special edition of the game to celebrate the Chinese Year of the Dog.

It’s actually pretty fun, and it comes with the self-satisfaction of knowing at least someone – or more accurately, something – can recognise your shitty sketches.

Bicycle.

But, of course, it doesn’t always work. For example, it failed to recognise the below drawing as a cat, despite my immaculate artistry:

Cat.

You can see the AI thinks my drawing most resembles an oven. Sure, it looks a bit like a teddy bear and hey, maybe the ears are a little small. Okay, the ears are actually terrible. Fine, the whole thing is just awful.

But let me get this straight: an OVEN!?

It even showed me what the AI thinks a cat looks like based on drawings by other people:

Other cats.

Give me a break! Some of these don’t even have faces, and the second-last sketch looks like some kind of science experiment gone horrifically wrong.

Go on, try it for yourself and see if you can do better.

There’s a cool visualisation of all the dog sketches that have been in the Chinese New Year version, which you can play here.

More than a billion doodles have been submitted to Quick, Draw! since its inception, so we can only assume it’s a few weeks from learning how to make and subsequently operate guns in order to enslave the human race.

In all seriousness, it’s a beautifully simple insight into how how machine learning is going to change the world we live in.

A post on Google’s official blog from December last year gives some pretty cool insights into the data.

For example, it’s noted that almost everyone draws a face with a smile; also, the perspective you use when drawing a chair seems to depend on where you’re from.

Quick, Draw! is only going to get better the more doodles it tries to guess, and all the data is shared publicly to help advance machine learning research.

About the author

Andrew is Techly’s Editor. Loves: weird gadgets and the Collingwood Football Club. Hates: olives and cardboard boxes.

Videos trending right now

In partnership with our friends at Glenmorangie, our sister site Lost At E Minor presents the third episode in the We Are The World We Create podcast series, dedicated to celebrating those people, and their exceptional creations, that help make the world a more interesting place. In this episode, Bigsy speaks to Masterchef contestant Reynold Poernomo and street artist Brad “Beastman” Eastman, delving into why dessert is everyone’s favourite course (and how you can make a living out of it) and how being a street artist can change how you perceive the world. Listen now!

Leave a comment