Facebook has recently revealed its plans with artificial intelligence (AI). The company has explained how it utilizes user data in order to compete with AI inventions from Google and Amazon. According to Wired, Facebook researchers are using 3.5 billion Instagram photos, sourced from 17,000 hashtags. The researchers are using the posts to train AI to tell images apart from each other. If the innovation is successful, people will not need to manually label images. As Wired notes, the project is 10 times the size of Google’s image-based experiments.

Facebook claims that out of 1 billion Instagram posts, its AI could correctly identify 85% of images. This beats Google’s attempt, which could identify 83%. The posts that are easiest to identify are #dogs, #cats, and #sunsets. Using this technology, Facebook hopes that it can remove violent and extremist content from its platform. The company already uses tools to remove nude and violent images. However, some claim the tools aren’t always consistent and effective.

Just remember next time you go to Instagram your pet, you may be training Facebook’s artificial intelligence experiment.