Anyone with a Web camera, Internet access, and a bit of creativity can train a machine (via Google)
Google's latest AI Experiment, Teachable Machine, is a teachable moment.
Anyone with a Web camera, Internet access, and a bit of creativity can train a machine-no coding required.
Fire up the program in your browser (desktop or mobile), then play around with different responses; a tutorial walks new users through the steps.
Start by looking into the camera to record input in three different classes: green, purple, orange. While holding the colored button (to capture at least 30 examples), try waving your hand, making a funny face, and staring blankly.
Then, when you recreate each action, the machine will respond with the corresponding GIF, sound, or robotic phrase. Try different approaches-even confuse the system on purpose-while you, and it, learn.
"We wanted to make it easier for people who are curious about this technology to learn more about it," Barron Webster, Google Creative Lab designer and machine-learning enthusiast, wrote in a blog announcement.
A collaborative effort by Copenhagen-based Stoj, LA's Use All Five, and the Creative Lab and PAIR teams at Google, the experiment was built using the deeplearn.js library.
Designed for everyone-from novice to master-the program makes it easy for Web developers to dive into machine learning by training and running neural networks directly in the browser.
Earlier this year, Google introduced AI Experiment AutoDraw, a Web-based tool that pairs machine learning with artist-created designs to help users better express themselves.
Simply sketch your best rendition of a cat, pizza, houseboat, or rabbit's foot keychain, and the computer will fill in the blanks-like an arts-and-crafts version of autocorrect.
The suggestion tool uses the same technology as Google's Quick, Draw!, another of the company's growing AI Experiments intended to prove artificial intelligence can be applied to human pursuits.