2 min read
Named ‘I-JEPA’, the AI tool uses background knowledge about the world inorder to fill missing pieces of images, an approach that is different from regular generative AI models
Meta, the parent company to Facebook and Instagram has just announced that it will provide researchers with a new ‘human-like’ AI image creation model that can analyze and complete unfinished parts of an image.
Named as the ‘I-JEPA’, the AI tool uses background knowledge about the world inorder to fill in missing pieces of images. While the I-JEPA might sound like a regular AI image generation and fill tool, it is far different and is claimed to be a lot better.
What makes the I-JEPA different is the fact that it uses background knowledge about the world when filling in images, making the output a lot better in terms of imagination and does not limit outcomes to the fill. Regular image filling AI tools, on the other hand, analyze the surrounding pixels of an incomplete image and complete the photo based on the data they collect from the image itself.
Meta says that being imaginative and having prior knowledge makes their AI tool to be a lot more ‘human-like’. Yann LeCun, a Meta AI scientist says that being human-like, allows the AI to avoid common AI mistakes such as generating a hand with extra fingers etc.
“For us, it’s way better if the industry standardizes on the basic tools that we’re using and therefore we can benefit from the improvements that others make,” said LeCun.
Placing a lot of its bets on artificial intelligence, Meta is continually releasing AI models or even AI related models. Just recently, Meta incorporated generative AI features into its ad tools, while also announcing an Instagram tool that will edit user photos based on their text commands.
Read more:
