Meta Launches Human-Like AI Image Regeneration Model for Facebook and Instagram

The image regeneration AI model by Meta through research will outdo existing artificial intelligence technology, by creating sharp images based on text prompts.

Written by: Arjita Gupta

Edited by: Krutarth Ghosalkar

Last Updated: June 14, 2023

Mark Zuckerberg, the chief executive of Meta has commented that Meta researchers are helping the firm to spur innovation and share models on AI and that the Meta platform will offer the researchers access to new ‘human-like’ AI model components, which can complete and analyze unfinished images with precision than previous models. The AI model, I-JEPA will be able to fill the missing image pieces by using background knowledge about the world.

It will not look into just nearby pixels like the existing AI models, which otherwise generate such images, Meta said. Meta owns Instagram and Facebook. What’s new with their AI model is that the approach is more human-like and incorporates reasoning based on Yann LeCun, the top AI scientist of Meta. It also has hi-tech methods to shirk away from errors commonly seen in artificial intelligence-based images such as hands with additional fingers.

Meta is currently implementing AI features in ad tools and other consumer products to create image backgrounds. And on Instagram, artificial intelligence technology based on text prompts can modify user photos. The company’s in-house research lab is a prolific publisher of research on open-sourced AI. Mark Zuckerberg, the CEO has stated that sharing models which Meta researchers develop can assist the organization to lower costs and spotting safety gaps through innovation.

He also said that if the industry makes it the norm to use basic tools the company is using, it becomes way better for them, as they can benefit from whichever improvements others make. But the executives from the organization have not paid heed to the warnings from industry experts about the ill effects of utilizing AI technology.

They have not signed a statement backed by leaders from DeepMind, Google, Microsoft, and OpenAI last month, which equaled the risk of AI with wars and pandemics. Even Lecun the iconic godfather of AI has argued to build focused safety checks into artificial intelligence systems and is campaigning against AI doomerism.