"Machine learning" and "neural network" are common terms for anyone who follows what Google does today, but may not be the most accessible or understandable concepts for the masses. And that's fine – you do not have to master machine learning, for example, to enjoy better photos or keyboards. Still, Google has quietly demonstrated ways in which users can get to know these concepts better, and the latest experiment is a game called Emoji Scavenger Hunt.
Here's how it works: Once you open the game in your mobile browser and tap "Let's Play," you'll be asked to grant access to your phone's camera (privacy-conscious browsers such as Brave or Firefox Focus cooperate Not). Once you allow access to the camera, you'll see a countdown and hear some cartoon-like sounds (if you're not louder), then you'll be shown an emoji to find around you in the real world timer strikes zero. With each emoji you see IRL, more seconds will be added.
The experimental game has a few quirks, of course: On the one hand, object identification is far from perfect. In the few rounds I played, false positives were common. A rather joyless sister of the Google Assistant tells your search for Emoji and every few seconds asks aloud if the identification of the currently visible object is correct. If you focus on a single object throughout the turn, it will endlessly ask if that is it [object].
Emoji Scavenger Hunt is amusing and worth a turn, but it's not meant well polished game for mainstream game. Google's goal is to show "how machine learning can be used in a fun way", and the player here is not just you ̵
The experiment is powered by TensorFlow and does not access any servers, so Google does not capture or store any content viewed with your camera during gameplay. For those interested in Emoji Scavenger Hunt, the open source code is available on GitHub.
This is part of Google's AI Experiment Showcase, which makes machine learning more accessible through simple visual and auditory experiences.