Meta Platforms unveiled plans to provide researchers with access to a new “humanlike” artificial intelligence (AI) model that can analyze and complete images with a higher degree of accuracy than existing models.
The tech giant noted that the new model (i-JEPA) deploys prior knowledge about the world to fill in missing parts of images, rather than just looking at nearby components like other generative AI models.
Backed by the chief scientist of AI, Jan Lucan, this approach helps the technology avoid common errors in images generated by AI, such as the appearance of a human hand with extra fingers.
Chief Executive Mark Zuckerberg underlined that sharing models developed by Meta researchers can help the company by spurring innovation, discovering security vulnerabilities, and reducing costs.
“For us, it’s best if the industry standardizes on what we use in terms of core tools, so we can benefit from the improvements that others are making,” he told investors in April.
In the same vein, Meta started integrating features of generative AI models into its consumer products, including advertising tools that can create photo backgrounds and an Instagram product that can edit photos, both based on user searches.