Meta Platforms (META.O), the parent company of Facebook and Instagram, announced on June 13 that it will provide researchers with access to components of a new AI model called I-JEPA. According to Meta, I-JEPA is a “human-like” AI model that excels at analyzing and completing unfinished images with greater accuracy compared to existing models.
Unlike other generative AI models that focus on nearby pixels, I-JEPA utilizes background knowledge about the world to fill in missing parts of images. This approach aligns with the human-like reasoning advocated by Meta’s top AI scientist, Yann LeCun, and helps prevent common errors found in AI-generated images, such as hands with extra fingers.
Meta has a strong commitment to open-source AI research through its in-house research lab. CEO Mark Zuckerberg believes that sharing models developed by Meta’s researchers can drive innovation, identify safety gaps, and reduce costs for the company. Zuckerberg expressed his preference for industry standardization on the tools used by Meta, enabling them to benefit from improvements made by others.
Despite concerns raised by industry peers, Meta’s executives have dismissed warnings about the potential dangers of the technology. They declined to sign a statement supported by top executives from OpenAI, DeepMind, Microsoft (MSFT.O), and Google (GOOGL.O), which likened the risks associated with AI to pandemics and wars.
LeCun, often regarded as one of the pioneers of AI, has criticized “AI doomerism” and advocated for the implementation of safety checks in AI systems.
Meta is also integrating generative AI capabilities into its consumer products. For instance, it is developing ad tools that can create image backgrounds and an Instagram product that can modify user photos based on text prompts.