Meta Platforms (META.O) announced on Tuesday that it would supply researchers with components of a new “human-like” artificial intelligence model that can evaluate and complete incomplete photos more accurately than existing models.
The business stated I-JEPA leverages past knowledge about the world to fill in missing images, unlike other generative AI models that merely look at surrounding pixels.
According to Meta’s top AI scientist Yann LeCun, this solution avoids common AI-generated image problems like extra-fingered hands.
Facebook and Instagram owner Meta’s research lab publishes open-sourced AI research. Mark Zuckerberg, CEO of Facebook, says sharing Meta’s researchers’ models can boost innovation, identify safety issues, and reduce expenses.
“For us, it’s way better if the industry standardizes on the basic tools that we’re using and therefore we can benefit from the improvements that others make,” he told investors in April.
Last month, top executives from OpenAI, DeepMind, Microsoft (MSFT.O), and Google (GOOGL.O) compared the technology’s hazards to pandemics and wars, but the company’s executives refused to sign.
Lecun, one of the “godfathers of AI,” has opposed “AI doomerism” and advocated for AI safety measures.
Meta adds abundant AI elements to its consumer products, like ad tools that produce image backgrounds and an Instagram product that edits user photos based on text prompts.