Meta Platforms, the parent company of Facebook and Instagram, announced on Tuesday its plans to grant researchers access to components of its new "human-like" artificial intelligence (AI) model. This advanced model, named I-JEPA, boasts superior accuracy in analysing and completing unfinished images compared to existing models.
Unlike other generative AI models that primarily focus on neighbouring pixels, I-JEPA utilises background knowledge about the world to fill in the missing portions of images. This unique approach aligns with the human-like reasoning advocated by Meta's esteemed AI scientist, Yann LeCun, and helps mitigate common errors often found in AI-generated images, such as hands with extra fingers.
As a renowned publisher of open-source AI research through its in-house research lab, Meta believes that sharing models developed by its researchers contributes to fostering innovation, identifying safety vulnerabilities, and reducing costs. Meta's CEO, Mark Zuckerberg, highlighted the benefits of industry standardisation on the tools they employ, allowing them to capitalise on enhancements made by others. In April, he emphasised this viewpoint during an investor discussion.
Despite concerns voiced by industry peers about the potential risks associated with AI, Meta's executives have chosen not to sign a statement endorsed by top executives from OpenAI, DeepMind, Microsoft, and Google. This statement drew parallels between the risks of AI and those of pandemics and wars.
LeCun, widely recognised as one of the "godfathers of AI," actively opposes the notion of "AI doomerism" and advocates for the incorporation of safety checks within AI systems.
Also Read
'Buying Netflix at $4 billion would've been better instead of...': Former Yahoo CEO Marissa Mayer
ChatGPT beats top investment funds in stock-picking experiment