Meta Platforms announced it will grant researchers access to components of its new “human-like” artificial intelligence model called I-JEPA. This model has shown superior accuracy in analyzing and completing unfinished images compared to existing models, according to Meta.
Unlike other generative AI models that rely solely on nearby pixels, I-JEPA leverages background knowledge about the world to fill in missing elements of images. This approach aligns with the human-like reasoning advocated by Meta’s top AI scientist, Yann LeCun, and helps the technology avoid common errors found in AI-generated images, such as the inclusion of extra fingers on hands.
As the parent company of Facebook and Instagram, Meta has established itself as an active publisher of open-source AI research through its in-house research lab. CEO Mark Zuckerberg has emphasized the benefits of sharing models developed by Meta’s researchers, including fostering innovation, identifying safety gaps, and reducing costs.
Mark Zuckerberg expressed his belief that it is advantageous for the industry to standardize on the foundational tools used by Meta. By doing so, the company can benefit from the improvements made by others. He conveyed this perspective to investors in April.
Despite concerns raised by industry peers regarding the potential risks associated with AI, Meta’s executives have downplayed these warnings. They declined to sign a statement, supported by top executives from OpenAI, DeepMind, Microsoft, and Google, that equated the risks of AI with those of pandemics and wars.
LeCun, regarded as one of the pioneers of AI, has actively opposed “AI doomerism” and advocated for the integration of safety checks within AI systems.
Meta is also integrating generative AI capabilities into its consumer products. For instance, they are developing ad tools that can create image backgrounds and an Instagram feature that can modify user photos based on text prompts.
