Google has updated its Google Photos app on Android with a new option that allows users to tell search giants about the content of their pictures. By labeling these images, Google can improve its object budget recognition algorithms, making photos more useful. It is a virtuous cycle of AI development that is best deployed by tech giants like Google that have a lot of data and a lot of users.
This is not an unusual practice. Machine learning systems don’t just learn by themselves, and most of these applications need to be taught using data labeled by humans. That’s why CAPTCHA tells you to identify cars and motorbikes in images. You are training AI to do this by recognizing these objects and objects.
The feature appears in the most recent version of Google Photos. Just tap the search button in the app’s menu, scroll down and you’ll see the “Help improve Google Photos” option. Reported by 9to5Google, Click on it and you will be presented with four functions: to describe your printing preferences for photos; Your chosen collage or animation; To identify which photos belong to which holiday events (e.g. Christmas or Halloween); And to identify the content of photos (“Name the most important items in this photo”).
:format(webp):no_upscale()/cdn.vox-cdn.com/uploads/chorus_asset/file/22028428/google_photos_ai.jpg)
As Google explains on the Help page about the feature: “It will take time to see if your contributions fall into your account, but your input will help improve and update existing features; For example, improved suggestions on how to print photos or high quality creations that you like. You can delete your answers at any time. “(To do so, tap the three-dot menu at the top right of the screen and press” Delete My Answers. “) As you type, the update appears to be available on Android, not just iOS.
Although this seems like a new addition to the Google Photos app, the underlying software is very old. The process is powered by “Crowdsource by Google,” a crowdfunding platform that the company launched in 2016. This introduces data labeling. Positive or negative, for example), transcription of handwritten notes and other similar jobs. Clearly, users don’t get any real rewards from Google for their work beyond virtual kudos.
All of this is worth remembering when using Google’s Vizi Machine Learning Products: they won’t be half as good without helping humans teach them.
