Google wants to make AI smarter with your bad doodles

16 Nov 2016

Doodle drawing. Image: igorstevanovic/Shutterstock

Unlike the beautifully designed series of Google Doodles, a new game released by the company is reversing the concept, calling on budding artists to draw their best doodles.

With so many major tech companies trying to improve their machine-learning AI these days, the internet has become awash with different fun games designed to get the public to chip in, helping to make their systems that bit smarter.

Now as part of its latest release of experiments into AI, Google has unveiled a new online game called Quick, Draw!

Future Human

Crowdsourcing bad doodles

The game is as simple as you can imagine, at least in theory, as the player is presented with a series of words, and they are then asked to draw a representation of that image in under 20 seconds.

Aside from it making you painfully aware of your lack of artistic skills, the purpose of the game is to help the AI crowdsource a number of different variations of one object, to learn the key characteristics that make the item distinct from others.

As you start to draw, Google’s AI will begin shouting words at you as to what it thinks it is, which might cause a bit of frustration. While you might think that is the best representation of a basket you could possibly create, Google might not see it the same way.

After you’ve played the game, Google will then tell you how your drawings compared with other people’s, and what exactly the AI thought your drawings were.

Traffic light doodle

Quick, Draw! telling you what it thought you were trying to draw. Image: Quick, Draw!/Google

New Cloud Machine Learning group

A number of different experiments have been included in Google’s latest batch of AI experiments which (aside from Quick, Draw!) includes a number of mobile-based games, such as the Thing Translator that lets you take a picture of something to hear how to say it in a different language.

Today’s release coincides with the company making a number of announcements surrounding AI, and where it plans to take the technology over the coming years.

One of the first major moves was the establishment of a new Cloud Machine Learning group led by Snapchat’s former head of research, Jia Li, and the former head of Stanford’s AI Lab, Fei-Fei Li.

In a blog post, the group lead for Google Cloud Machine Learning, Rob Craft, said of the new group: “Building a centralised team within Google Cloud will accelerate our ability to deliver machine-learning products and services to enterprise customers in every industry.”

Colm Gorey was a senior journalist with Silicon Republic